diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..20b406e63a1abf865fd144a322b34165c30817ec
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-100/trainer_state.json
@@ -0,0 +1,139 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.24906600249066002,
+  "eval_steps": 20,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9823576763965440.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfe73a5743fcfda8d7ec364b883e78080a8ec471
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1000/trainer_state.json
@@ -0,0 +1,1084 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.488169364881694,
+  "eval_steps": 20,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.859037950771814e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c462a907d1cc6701c162efe12def24544e8d019
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1020/trainer_state.json
@@ -0,0 +1,1105 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.5379825653798256,
+  "eval_steps": 20,
+  "global_step": 1020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0076952699436032e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..907ef505bb00b812dd0a2e31186361269113de00
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1040/trainer_state.json
@@ -0,0 +1,1126 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.587795765877958,
+  "eval_steps": 20,
+  "global_step": 1040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0254458345271091e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ec0e8941bdb4365303d4efa80f64f3787def7ee
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1060/trainer_state.json
@@ -0,0 +1,1147 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.6376089663760895,
+  "eval_steps": 20,
+  "global_step": 1060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.045743734380032e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..51cf569b8933e96ea4343c3d0f1ea233af51977f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1080/trainer_state.json
@@ -0,0 +1,1168 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.6874221668742218,
+  "eval_steps": 20,
+  "global_step": 1080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0682640451304448e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a85b886556ea5dcc6d8507b8b863b00d9ecbd9ad
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1100/trainer_state.json
@@ -0,0 +1,1189 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.7372353673723535,
+  "eval_steps": 20,
+  "global_step": 1100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0877774590688256e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5b41f15a345ccc37f07a171ccd3f0f644d342bc
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1120/trainer_state.json
@@ -0,0 +1,1210 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.7870485678704857,
+  "eval_steps": 20,
+  "global_step": 1120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1058529480242586e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..50cc56468ed358108ec186f62f5a4ab62eeba058
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1140/trainer_state.json
@@ -0,0 +1,1231 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8368617683686175,
+  "eval_steps": 20,
+  "global_step": 1140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1277422592347136e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e70755e11308cfb4149a322fc6f55d2ef901a1b4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1160/trainer_state.json
@@ -0,0 +1,1252 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8866749688667497,
+  "eval_steps": 20,
+  "global_step": 1160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1472790102826803e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d0e7c0d0b440270b68b490316b86c1f3911c717
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1180/trainer_state.json
@@ -0,0 +1,1273 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.936488169364882,
+  "eval_steps": 20,
+  "global_step": 1180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1675715246487757e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9cc6b9812fd08b33630c2d416efe72e5b1ff015
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-120/trainer_state.json
@@ -0,0 +1,160 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.298879202988792,
+  "eval_steps": 20,
+  "global_step": 120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1745836528934912e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..59934fc7363f00f56f46062daaca05db1ad179c9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1200/trainer_state.json
@@ -0,0 +1,1294 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9863013698630136,
+  "eval_steps": 20,
+  "global_step": 1200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1868291279628493e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..95a5ab0ac38c50b53455bdfd6811db6319d068a8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1220/trainer_state.json
@@ -0,0 +1,1315 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0348692403486925,
+  "eval_steps": 20,
+  "global_step": 1220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2056756391157555e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aec7537030240eed6c20331d13391fab738fd742
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1240/trainer_state.json
@@ -0,0 +1,1336 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0846824408468243,
+  "eval_steps": 20,
+  "global_step": 1240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2245230478497997e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e4a9e65ddd7141054bceeed5b49dade599d2bdb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1260/trainer_state.json
@@ -0,0 +1,1357 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.1344956413449565,
+  "eval_steps": 20,
+  "global_step": 1260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2456036384549888e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..63d194660b1f81844845f4d4ea8ddb8679158547
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1280/trainer_state.json
@@ -0,0 +1,1378 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.1843088418430883,
+  "eval_steps": 20,
+  "global_step": 1280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.26603168757291e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a8867bdc001082cf64129bfa6b0abc24528b0b7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1300/trainer_state.json
@@ -0,0 +1,1399 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2341220423412205,
+  "eval_steps": 20,
+  "global_step": 1300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2852129964902605e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bbd5eb00ae283dad8d104a0bcfd6e8a8688d989
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1320/trainer_state.json
@@ -0,0 +1,1420 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2839352428393527,
+  "eval_steps": 20,
+  "global_step": 1320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3068070035063398e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d5600a352f4805edc32868811fa1026047cadf3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1340/trainer_state.json
@@ -0,0 +1,1441 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.3337484433374844,
+  "eval_steps": 20,
+  "global_step": 1340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3249722505755648e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..24b3460e6fdaaaa48012ab4783d7576f7591c515
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1360/trainer_state.json
@@ -0,0 +1,1462 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.383561643835616,
+  "eval_steps": 20,
+  "global_step": 1360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3445125919480832e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4aae3db6d6d3767d6090d67b91f26c133804e54
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1380/trainer_state.json
@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.4333748443337484,
+  "eval_steps": 20,
+  "global_step": 1380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.363284603866542e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..00104806855881c2ec3c2ddda8b5c7e4c9252d85
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-140/trainer_state.json
@@ -0,0 +1,181 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.34869240348692404,
+  "eval_steps": 20,
+  "global_step": 140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3754802631827456e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dbc57c0b1e656d3a3fb257bc062c4cc4ec6df40
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1400/trainer_state.json
@@ -0,0 +1,1504 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.4831880448318806,
+  "eval_steps": 20,
+  "global_step": 1400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.385097620680274e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..91337e1cd619daa47e3d0280def4011f7b9b2f72
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1420/trainer_state.json
@@ -0,0 +1,1525 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.5330012453300124,
+  "eval_steps": 20,
+  "global_step": 1420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.405775197354537e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f167215a10c9e913263cbf92252f420701084cee
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1440/trainer_state.json
@@ -0,0 +1,1546 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.5828144458281446,
+  "eval_steps": 20,
+  "global_step": 1440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4260650189772186e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..70abf49a52c47bf637b67c73ed750c68fab508cc
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1460/trainer_state.json
@@ -0,0 +1,1567 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.6326276463262763,
+  "eval_steps": 20,
+  "global_step": 1460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4456080530931507e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..21da5b2b10015a69367d980a1a4c57ed267533b8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1480/trainer_state.json
@@ -0,0 +1,1588 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.6824408468244085,
+  "eval_steps": 20,
+  "global_step": 1480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4642463254220595e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a91dedb8d575326148b8f962f7e6e2f5c50a270
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1500/trainer_state.json
@@ -0,0 +1,1609 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.7322540473225407,
+  "eval_steps": 20,
+  "global_step": 1500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4841169766533325e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a926984fa869df7e00d8aba8cccc3a031640fae
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1520/trainer_state.json
@@ -0,0 +1,1630 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.7820672478206725,
+  "eval_steps": 20,
+  "global_step": 1520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5019420404712858e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d1c3f01f4598d1ad6cb2277961cd1fd75d0093f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1540/trainer_state.json
@@ -0,0 +1,1651 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.8318804483188043,
+  "eval_steps": 20,
+  "global_step": 1540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5214482737605632e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2afa0a884859ed868bec7faba2e0a37236638669
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1560/trainer_state.json
@@ -0,0 +1,1672 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.8816936488169365,
+  "eval_steps": 20,
+  "global_step": 1560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.541802721225175e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..17d4fb82022c168b921d22b7f2f50b18f8beceba
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1580/trainer_state.json
@@ -0,0 +1,1693 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9315068493150687,
+  "eval_steps": 20,
+  "global_step": 1580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.559732802036265e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f51eff8583228db5644e4fee421cc409ded8fb3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-160/trainer_state.json
@@ -0,0 +1,202 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.398505603985056,
+  "eval_steps": 20,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5751830905585664e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c2b2227185da1b5072ed938eaaa9d6750146871
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1600/trainer_state.json
@@ -0,0 +1,1714 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9813200498132004,
+  "eval_steps": 20,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5786448366122394e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..05fca03dc0e75e1bd3d161e2d40d2d62b135789e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1620/trainer_state.json
@@ -0,0 +1,1735 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.029887920298879,
+  "eval_steps": 20,
+  "global_step": 1620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5984833993178317e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..50fdc814b097d602100d4d4f86f018b8bd256ae6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1640/trainer_state.json
@@ -0,0 +1,1756 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0797011207970115,
+  "eval_steps": 20,
+  "global_step": 1640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6179941205127987e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1907262b85ce5386b1de634b9df85698f044660e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1660/trainer_state.json
@@ -0,0 +1,1777 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.129514321295143,
+  "eval_steps": 20,
+  "global_step": 1660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6397263550241178e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d919d58536c3b515e173dc864e76744b4e1d3a2c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1680/trainer_state.json
@@ -0,0 +1,1798 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.179327521793275,
+  "eval_steps": 20,
+  "global_step": 1680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.659570078821253e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc1d7781bbcff7d9e2166e6c1a1e45fd514e2597
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1700/trainer_state.json
@@ -0,0 +1,1819 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.229140722291407,
+  "eval_steps": 20,
+  "global_step": 1700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.67823079067861e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca850995f6bdb5f6032d08745a0f118e386bc043
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1720/trainer_state.json
@@ -0,0 +1,1840 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.2789539227895395,
+  "eval_steps": 20,
+  "global_step": 1720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6969022735096218e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b48dd1820a14b1fb6df848ac327194c1d69e5e1c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1740/trainer_state.json
@@ -0,0 +1,1861 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.328767123287671,
+  "eval_steps": 20,
+  "global_step": 1740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7173159613293363e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bbf014cafc49db662e46901feee5a2b2c74017d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1760/trainer_state.json
@@ -0,0 +1,1882 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.378580323785803,
+  "eval_steps": 20,
+  "global_step": 1760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7368688688377856e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7befad6c9ef53e98e60eda24e6c2fbbf36adb411
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1780/trainer_state.json
@@ -0,0 +1,1903 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.428393524283935,
+  "eval_steps": 20,
+  "global_step": 1780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.756396644074373e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a850eac6dc5393e98ab6b17ddcb0846152a45f5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-180/trainer_state.json
@@ -0,0 +1,223 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.44831880448318806,
+  "eval_steps": 20,
+  "global_step": 180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7813036230705152e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..640abe3427039b775c214983af5c2871db966c17
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1800/trainer_state.json
@@ -0,0 +1,1924 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.478206724782067,
+  "eval_steps": 20,
+  "global_step": 1800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7752727754048307e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac72acebba6fa533c2a50c885055bd74c7d21f2e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1820/trainer_state.json
@@ -0,0 +1,1945 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.5280199252802,
+  "eval_steps": 20,
+  "global_step": 1820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7933742942132634e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..efced28cc370833c69672a34874e350d19510466
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1840/trainer_state.json
@@ -0,0 +1,1966 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.577833125778331,
+  "eval_steps": 20,
+  "global_step": 1840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8121184811164467e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0d138ae627cba49a4d02b3e8afbd53148b48d53
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1860/trainer_state.json
@@ -0,0 +1,1987 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.627646326276463,
+  "eval_steps": 20,
+  "global_step": 1860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.83182936290517e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c62e30cbade6f04a7f30343b86f2fa7f1d51dc9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1880/trainer_state.json
@@ -0,0 +1,2008 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.677459526774595,
+  "eval_steps": 20,
+  "global_step": 1880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.85193428281344e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..61871064ac4543fa53e0a481622ca2212080f0d3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1900/trainer_state.json
@@ -0,0 +1,2029 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.7272727272727275,
+  "eval_steps": 20,
+  "global_step": 1900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.87351931401814e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4de2866a0813dcb525d96866fbfa9e1a0b6dda48
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1920/trainer_state.json
@@ -0,0 +1,2050 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.777085927770859,
+  "eval_steps": 20,
+  "global_step": 1920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.894537073943675e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c471080274e2f83f0f4c13e816a6fa87e093323b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1940/trainer_state.json
@@ -0,0 +1,2071 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.826899128268991,
+  "eval_steps": 20,
+  "global_step": 1940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.913542456957993e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..72792742464fb5f9c47c9546b381a170eb274a78
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1960/trainer_state.json
@@ -0,0 +1,2092 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.876712328767123,
+  "eval_steps": 20,
+  "global_step": 1960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9335782631186432e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac7e78a1dcbe093d32c86a5e1a75859fff555268
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-1980/trainer_state.json
@@ -0,0 +1,2113 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.926525529265255,
+  "eval_steps": 20,
+  "global_step": 1980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9516824746704896e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..91e972f4341a24861abe46be31f13ed17be9832b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3460/trainer_state.json
@@ -0,0 +1,3667 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.607721046077211,
+  "eval_steps": 20,
+  "global_step": 3460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4093681551602074e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5584241ec6b9c566311bb39a43f0b2889d3a405f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3480/trainer_state.json
@@ -0,0 +1,3688 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.657534246575342,
+  "eval_steps": 20,
+  "global_step": 3480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4298419809161626e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9663eb44b248758b63dc792d36122b642b56113
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3500/trainer_state.json
@@ -0,0 +1,3709 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.707347447073474,
+  "eval_steps": 20,
+  "global_step": 3500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4516307630391706e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d77329a279c6d01c0873cfd431d5c144114c73a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3520/trainer_state.json
@@ -0,0 +1,3730 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.757160647571606,
+  "eval_steps": 20,
+  "global_step": 3520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.471473589255168e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbacf3484769c514d268b015c46536a15fad5ec0
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3540/trainer_state.json
@@ -0,0 +1,3751 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.806973848069738,
+  "eval_steps": 20,
+  "global_step": 3540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4923127305342566e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2cb49980d844e1bc4a83cad951e198314aca95c4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3560/trainer_state.json
@@ -0,0 +1,3772 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.85678704856787,
+  "eval_steps": 20,
+  "global_step": 3560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.509752732044042e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfbb82c3a37c503f8111d7c528df3fb3783af438
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3580/trainer_state.json
@@ -0,0 +1,3793 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.906600249066003,
+  "eval_steps": 20,
+  "global_step": 3580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.527526633737134e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0db24247f8c8234f2fe3786eb5c30cc503e3b1b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-360/trainer_state.json
@@ -0,0 +1,412 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8966376089663761,
+  "eval_steps": 20,
+  "global_step": 360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5892743817486336e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..517815cb4b55e0e5f06c8ed6dff8a6232cf4242c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3600/trainer_state.json
@@ -0,0 +1,3814 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.956413449564135,
+  "eval_steps": 20,
+  "global_step": 3600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5487499397432525e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c74165afdccfd2fe75bc8e2d6f3ae1d915eab8a8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3620/trainer_state.json
@@ -0,0 +1,3835 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.004981320049813,
+  "eval_steps": 20,
+  "global_step": 3620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5680073186620416e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..da92f8a3cc8d5d17652922599e6a9560c7717414
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3640/trainer_state.json
@@ -0,0 +1,3856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.054794520547945,
+  "eval_steps": 20,
+  "global_step": 3640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.586990262147912e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cb2bd64c823b3818433455081a6279d50b85743
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3660/trainer_state.json
@@ -0,0 +1,3877 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.104607721046078,
+  "eval_steps": 20,
+  "global_step": 3660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.607405745129902e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..be6f5adccfc47dabe4fb481ef9f83a309ad84fda
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3680/trainer_state.json
@@ -0,0 +1,3898 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.15442092154421,
+  "eval_steps": 20,
+  "global_step": 3680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.626792600127836e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d46412d97cabcc8df1eaabdfa631cf39c46919a5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3700/trainer_state.json
@@ -0,0 +1,3919 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.204234122042342,
+  "eval_steps": 20,
+  "global_step": 3700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.6481290013573325e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6b1859f856f3191883f30ec802c9b60213392d3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3720/trainer_state.json
@@ -0,0 +1,3940 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.254047322540472,
+  "eval_steps": 20,
+  "global_step": 3720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.6676630596618854e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..24dd6c7932d1ed12cf45dc50955b9125d68e65de
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3740/trainer_state.json
@@ -0,0 +1,3961 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.303860523038605,
+  "eval_steps": 20,
+  "global_step": 3740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.686834495186719e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..926e772f3bc449ec3b4d4ae7fb6e4fb8619e0a9f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3760/trainer_state.json
@@ -0,0 +1,3982 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.353673723536737,
+  "eval_steps": 20,
+  "global_step": 3760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.7059314314771046e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..805cf0fa3fd3c7b64099156e5a7c2f4c426e6e7d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3780/trainer_state.json
@@ -0,0 +1,4003 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.403486924034869,
+  "eval_steps": 20,
+  "global_step": 3780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.725921460999721e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e77dba870644dcbbf2e0db9cc4cc05ae0857465
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-380/trainer_state.json
@@ -0,0 +1,433 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9464508094645081,
+  "eval_steps": 20,
+  "global_step": 380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.774840306202214e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..34d7b190d19766098a508daac13fe226fd05bfce
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3800/trainer_state.json
@@ -0,0 +1,4024 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.453300124533001,
+  "eval_steps": 20,
+  "global_step": 3800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.74812402802731e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..53f9f3117f84554f3478a67ddbace0bed1b147ff
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3820/trainer_state.json
@@ -0,0 +1,4045 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.503113325031133,
+  "eval_steps": 20,
+  "global_step": 3820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.767130308622766e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aced71fb07e6534e9340c70d5c351cfcb35f5df8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3840/trainer_state.json
@@ -0,0 +1,4066 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.552926525529266,
+  "eval_steps": 20,
+  "global_step": 3840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.785028076512891e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9efd58cef05d1dc74ffd362e2962acc81376577f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3860/trainer_state.json
@@ -0,0 +1,4087 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.602739726027398,
+  "eval_steps": 20,
+  "global_step": 3860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.804764090573476e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9717e2f7f0eb6f6b68099b17339ee4f9e942668c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3880/trainer_state.json
@@ -0,0 +1,4108 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.65255292652553,
+  "eval_steps": 20,
+  "global_step": 3880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.8251104598078464e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddce50db3e69fcf4ccfbf9fd6c466c6b898e443e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3900/trainer_state.json
@@ -0,0 +1,4129 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.70236612702366,
+  "eval_steps": 20,
+  "global_step": 3900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.845510683910492e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b7a0e528051bee6b418127244f0612bcc5c71e3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3920/trainer_state.json
@@ -0,0 +1,4150 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.752179327521793,
+  "eval_steps": 20,
+  "global_step": 3920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.8652538786201805e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..92891050e5b0ef9cff793212bee5605de8336ecb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3940/trainer_state.json
@@ -0,0 +1,4171 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.801992528019925,
+  "eval_steps": 20,
+  "global_step": 3940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    },
+    {
+      "entropy": 0.12030278495512903,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.20454250276088715,
+      "learning_rate": 2.797298117403634e-07,
+      "loss": 0.061210107803344724,
+      "mean_token_accuracy": 0.9774218738079071,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.2348609700105911,
+      "eval_loss": 0.9410302639007568,
+      "eval_mean_token_accuracy": 0.8523409498292346,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 85.9105,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 3940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.884385820574945e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..be74de236be8c225119e41c8504c18062294c49d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3960/trainer_state.json
@@ -0,0 +1,4192 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.851805728518057,
+  "eval_steps": 20,
+  "global_step": 3960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    },
+    {
+      "entropy": 0.12030278495512903,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.20454250276088715,
+      "learning_rate": 2.797298117403634e-07,
+      "loss": 0.061210107803344724,
+      "mean_token_accuracy": 0.9774218738079071,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.2348609700105911,
+      "eval_loss": 0.9410302639007568,
+      "eval_mean_token_accuracy": 0.8523409498292346,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 85.9105,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 3940
+    },
+    {
+      "entropy": 0.11162231937050819,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.13079790771007538,
+      "learning_rate": 1.5867404281932237e-07,
+      "loss": 0.05349419116973877,
+      "mean_token_accuracy": 0.9791230395436287,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.23484029950097549,
+      "eval_loss": 0.9411523342132568,
+      "eval_mean_token_accuracy": 0.8523351938225502,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 85.1298,
+      "eval_samples_per_second": 16.152,
+      "eval_steps_per_second": 2.02,
+      "step": 3960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.906717639286395e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce7180b7b8256b337de4fda1aea9f092b09530c8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-3980/trainer_state.json
@@ -0,0 +1,4213 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.90161892901619,
+  "eval_steps": 20,
+  "global_step": 3980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    },
+    {
+      "entropy": 0.12030278495512903,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.20454250276088715,
+      "learning_rate": 2.797298117403634e-07,
+      "loss": 0.061210107803344724,
+      "mean_token_accuracy": 0.9774218738079071,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.2348609700105911,
+      "eval_loss": 0.9410302639007568,
+      "eval_mean_token_accuracy": 0.8523409498292346,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 85.9105,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 3940
+    },
+    {
+      "entropy": 0.11162231937050819,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.13079790771007538,
+      "learning_rate": 1.5867404281932237e-07,
+      "loss": 0.05349419116973877,
+      "mean_token_accuracy": 0.9791230395436287,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.23484029950097549,
+      "eval_loss": 0.9411523342132568,
+      "eval_mean_token_accuracy": 0.8523351938225502,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 85.1298,
+      "eval_samples_per_second": 16.152,
+      "eval_steps_per_second": 2.02,
+      "step": 3960
+    },
+    {
+      "entropy": 0.12446248792111873,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705058217048645,
+      "learning_rate": 7.16918189283218e-08,
+      "loss": 0.061916273832321164,
+      "mean_token_accuracy": 0.9767685994505882,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.2347967088395773,
+      "eval_loss": 0.9411665201187134,
+      "eval_mean_token_accuracy": 0.8524297087691551,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.2026,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 3980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9249780299562394e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..510e50b0318f41db38c10ff49f55f9e68cd3840a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-40/trainer_state.json
@@ -0,0 +1,76 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.099626400996264,
+  "eval_steps": 20,
+  "global_step": 40,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3753325286326272.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb904f47ee935149ff36d61610b34fafa92bc622
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-400/trainer_state.json
@@ -0,0 +1,454 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9962640099626401,
+  "eval_steps": 20,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.964974918647808e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed0f3f1e65e0b73f457ca3f838d5667ca4a77ba0
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4000/trainer_state.json
@@ -0,0 +1,4234 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.951432129514322,
+  "eval_steps": 20,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    },
+    {
+      "entropy": 0.12030278495512903,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.20454250276088715,
+      "learning_rate": 2.797298117403634e-07,
+      "loss": 0.061210107803344724,
+      "mean_token_accuracy": 0.9774218738079071,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.2348609700105911,
+      "eval_loss": 0.9410302639007568,
+      "eval_mean_token_accuracy": 0.8523409498292346,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 85.9105,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 3940
+    },
+    {
+      "entropy": 0.11162231937050819,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.13079790771007538,
+      "learning_rate": 1.5867404281932237e-07,
+      "loss": 0.05349419116973877,
+      "mean_token_accuracy": 0.9791230395436287,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.23484029950097549,
+      "eval_loss": 0.9411523342132568,
+      "eval_mean_token_accuracy": 0.8523351938225502,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 85.1298,
+      "eval_samples_per_second": 16.152,
+      "eval_steps_per_second": 2.02,
+      "step": 3960
+    },
+    {
+      "entropy": 0.12446248792111873,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705058217048645,
+      "learning_rate": 7.16918189283218e-08,
+      "loss": 0.061916273832321164,
+      "mean_token_accuracy": 0.9767685994505882,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.2347967088395773,
+      "eval_loss": 0.9411665201187134,
+      "eval_mean_token_accuracy": 0.8524297087691551,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.2026,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 3980
+    },
+    {
+      "entropy": 0.13053828622214497,
+      "epoch": 9.951432129514322,
+      "grad_norm": 0.1882583051919937,
+      "learning_rate": 1.8809372719714117e-08,
+      "loss": 0.06384563446044922,
+      "mean_token_accuracy": 0.9753879077732563,
+      "num_tokens": 9341118.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.951432129514322,
+      "eval_entropy": 0.23490050339768098,
+      "eval_loss": 0.9411982297897339,
+      "eval_mean_token_accuracy": 0.8522821709860203,
+      "eval_num_tokens": 9341118.0,
+      "eval_runtime": 86.5264,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9449716498034074e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf9c6ce757e6c7dbed95225046881d1beb8ca4cb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-4020/trainer_state.json
@@ -0,0 +1,4255 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "eval_steps": 20,
+  "global_step": 4020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    },
+    {
+      "entropy": 0.48843890577554705,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.9082473516464233,
+      "learning_rate": 0.00021141500186075868,
+      "loss": 0.4309722423553467,
+      "mean_token_accuracy": 0.8686766296625137,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5543508351195691,
+      "eval_loss": 0.5478800535202026,
+      "eval_mean_token_accuracy": 0.8478029522784921,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.3835,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4777219031006098,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7448089122772217,
+      "learning_rate": 0.0002104266917731438,
+      "loss": 0.423325252532959,
+      "mean_token_accuracy": 0.8706337086856365,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.49857561550168106,
+      "eval_loss": 0.5511948466300964,
+      "eval_mean_token_accuracy": 0.8502220289651737,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.5399,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.988,
+      "step": 1020
+    },
+    {
+      "entropy": 0.4844174191355705,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.794029176235199,
+      "learning_rate": 0.00020940904126432,
+      "loss": 0.4176753044128418,
+      "mean_token_accuracy": 0.873535567522049,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.485467542222766,
+      "eval_loss": 0.5539286732673645,
+      "eval_mean_token_accuracy": 0.8495475081510322,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.135,
+      "eval_samples_per_second": 15.963,
+      "eval_steps_per_second": 1.997,
+      "step": 1040
+    },
+    {
+      "entropy": 0.49070929251611234,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.7558256983757019,
+      "learning_rate": 0.0002083623572438007,
+      "loss": 0.42867293357849123,
+      "mean_token_accuracy": 0.8696666076779366,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.490822730889154,
+      "eval_loss": 0.5434785485267639,
+      "eval_mean_token_accuracy": 0.850568296950917,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.4933,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 1060
+    },
+    {
+      "entropy": 0.47806114703416824,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6608979105949402,
+      "learning_rate": 0.00020728695537721047,
+      "loss": 0.4289727687835693,
+      "mean_token_accuracy": 0.8693130135536193,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.5285773256490397,
+      "eval_loss": 0.5444230437278748,
+      "eval_mean_token_accuracy": 0.8498796481032704,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.7091,
+      "eval_samples_per_second": 15.858,
+      "eval_steps_per_second": 1.984,
+      "step": 1080
+    },
+    {
+      "entropy": 0.5046216730028391,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.8428544998168945,
+      "learning_rate": 0.00020618315999108454,
+      "loss": 0.43131070137023925,
+      "mean_token_accuracy": 0.8701941035687923,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.49888394738352576,
+      "eval_loss": 0.5459766387939453,
+      "eval_mean_token_accuracy": 0.8511758872935938,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.2222,
+      "eval_samples_per_second": 15.947,
+      "eval_steps_per_second": 1.995,
+      "step": 1100
+    },
+    {
+      "entropy": 0.5212558470666409,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 1.129318118095398,
+      "learning_rate": 0.00020505130397505635,
+      "loss": 0.44249300956726073,
+      "mean_token_accuracy": 0.8654101334512234,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5179622324053631,
+      "eval_loss": 0.5522801280021667,
+      "eval_mean_token_accuracy": 0.8497019947268242,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.1903,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.996,
+      "step": 1120
+    },
+    {
+      "entropy": 0.4988406613469124,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6460545063018799,
+      "learning_rate": 0.00020389172868146263,
+      "loss": 0.4386270523071289,
+      "mean_token_accuracy": 0.8690383620560169,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.5042278484203094,
+      "eval_loss": 0.5433034300804138,
+      "eval_mean_token_accuracy": 0.8497674451317898,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.3028,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.993,
+      "step": 1140
+    },
+    {
+      "entropy": 0.4926559619605541,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8199329972267151,
+      "learning_rate": 0.00020270478382239615,
+      "loss": 0.4313485145568848,
+      "mean_token_accuracy": 0.8674727231264114,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.503873193160046,
+      "eval_loss": 0.5388111472129822,
+      "eval_mean_token_accuracy": 0.8526195034731266,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 86.4054,
+      "eval_samples_per_second": 15.913,
+      "eval_steps_per_second": 1.991,
+      "step": 1160
+    },
+    {
+      "entropy": 0.5020013231784105,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7344821095466614,
+      "learning_rate": 0.00020149082736423723,
+      "loss": 0.43590536117553713,
+      "mean_token_accuracy": 0.8671772189438343,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5368241809828337,
+      "eval_loss": 0.5355703830718994,
+      "eval_mean_token_accuracy": 0.8517617773871089,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.2945,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1180
+    },
+    {
+      "entropy": 0.5112275708466768,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6951606869697571,
+      "learning_rate": 0.00020025022541969622,
+      "loss": 0.43579301834106443,
+      "mean_token_accuracy": 0.8641206480562686,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.5066795706055885,
+      "eval_loss": 0.5415249466896057,
+      "eval_mean_token_accuracy": 0.8493563373421513,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.5005,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.988,
+      "step": 1200
+    },
+    {
+      "entropy": 0.42298635305502474,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.8201794028282166,
+      "learning_rate": 0.00019898335213739863,
+      "loss": 0.35593905448913576,
+      "mean_token_accuracy": 0.889238600547497,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.4584170470750609,
+      "eval_loss": 0.569487452507019,
+      "eval_mean_token_accuracy": 0.8495814173027526,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 86.2281,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 1220
+    },
+    {
+      "entropy": 0.37450140453875064,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7308394908905029,
+      "learning_rate": 0.0001976905895890471,
+      "loss": 0.307823920249939,
+      "mean_token_accuracy": 0.9001288741827012,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.45185995916294497,
+      "eval_loss": 0.5672881603240967,
+      "eval_mean_token_accuracy": 0.8511318519364955,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.0819,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3887945845723152,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.7299330830574036,
+      "learning_rate": 0.0001963723276541939,
+      "loss": 0.32047903537750244,
+      "mean_token_accuracy": 0.8960984498262405,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.44865354549053105,
+      "eval_loss": 0.5666037201881409,
+      "eval_mean_token_accuracy": 0.8496572649063066,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 1260
+    },
+    {
+      "entropy": 0.39677664265036583,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.9533219933509827,
+      "learning_rate": 0.00019502896390265838,
+      "loss": 0.3253983497619629,
+      "mean_token_accuracy": 0.8964207418262958,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.4641980809527774,
+      "eval_loss": 0.5814996957778931,
+      "eval_mean_token_accuracy": 0.8485886212005171,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.7784,
+      "eval_samples_per_second": 15.845,
+      "eval_steps_per_second": 1.982,
+      "step": 1280
+    },
+    {
+      "entropy": 0.39210722744464876,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7447651028633118,
+      "learning_rate": 0.00019366090347462545,
+      "loss": 0.3276803970336914,
+      "mean_token_accuracy": 0.8930055953562259,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43595615254585135,
+      "eval_loss": 0.5722188353538513,
+      "eval_mean_token_accuracy": 0.8501105755567551,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.5271,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3684127271175385,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 0.6934201121330261,
+      "learning_rate": 0.00019226855895846078,
+      "loss": 0.3156379222869873,
+      "mean_token_accuracy": 0.8976306475698947,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.4628148723480313,
+      "eval_loss": 0.5631352066993713,
+      "eval_mean_token_accuracy": 0.8504934813394103,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.3436,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 1320
+    },
+    {
+      "entropy": 0.4073401909321547,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.9386897683143616,
+      "learning_rate": 0.00019085235026627994,
+      "loss": 0.34265310764312745,
+      "mean_token_accuracy": 0.8902062118053437,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.46455050623694133,
+      "eval_loss": 0.5586736798286438,
+      "eval_mean_token_accuracy": 0.8506874702004499,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 86.1286,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 1340
+    },
+    {
+      "entropy": 0.4046429242938757,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9633992314338684,
+      "learning_rate": 0.00018941270450730836,
+      "loss": 0.33816893100738527,
+      "mean_token_accuracy": 0.8927541889250279,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.46846531660750856,
+      "eval_loss": 0.561501681804657,
+      "eval_mean_token_accuracy": 0.8496256377114806,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 86.685,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39872407019138334,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7786458730697632,
+      "learning_rate": 0.00018795005585907113,
+      "loss": 0.33342490196228025,
+      "mean_token_accuracy": 0.8944805048406124,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.42709505973860273,
+      "eval_loss": 0.5751848220825195,
+      "eval_mean_token_accuracy": 0.8507290447867194,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.6892,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 1380
+    },
+    {
+      "entropy": 0.3923338124528527,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.9305956363677979,
+      "learning_rate": 0.0001864648454364511,
+      "loss": 0.33188116550445557,
+      "mean_token_accuracy": 0.8943330392241478,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.4386174779298694,
+      "eval_loss": 0.5680831074714661,
+      "eval_mean_token_accuracy": 0.8513129727784977,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 86.2671,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3856233984231949,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 1.0362752676010132,
+      "learning_rate": 0.0001849575211586545,
+      "loss": 0.33098697662353516,
+      "mean_token_accuracy": 0.8961390435695649,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4574795474493226,
+      "eval_loss": 0.5630439519882202,
+      "eval_mean_token_accuracy": 0.8520988873964133,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.6035,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39812871962785723,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7807195782661438,
+      "learning_rate": 0.0001834285376141247,
+      "loss": 0.3333771228790283,
+      "mean_token_accuracy": 0.8930827379226685,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.4556825893909432,
+      "eval_loss": 0.5689062476158142,
+      "eval_mean_token_accuracy": 0.8507103507601937,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.1606,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.996,
+      "step": 1440
+    },
+    {
+      "entropy": 0.4147744856774807,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6429352164268494,
+      "learning_rate": 0.00018187835592344443,
+      "loss": 0.3482560873031616,
+      "mean_token_accuracy": 0.8910200245678425,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.46600024540757023,
+      "eval_loss": 0.5609709024429321,
+      "eval_mean_token_accuracy": 0.8491220876227977,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 86.8039,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40425071083009245,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.8613698482513428,
+      "learning_rate": 0.0001803074436002682,
+      "loss": 0.342916464805603,
+      "mean_token_accuracy": 0.8916418336331844,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.43855057899342026,
+      "eval_loss": 0.5720968246459961,
+      "eval_mean_token_accuracy": 0.8500823641932288,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 86.6746,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1480
+    },
+    {
+      "entropy": 0.39465143866837027,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6285189986228943,
+      "learning_rate": 0.0001787162744103265,
+      "loss": 0.3424591779708862,
+      "mean_token_accuracy": 0.8906558901071548,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4509461877304454,
+      "eval_loss": 0.5590082406997681,
+      "eval_mean_token_accuracy": 0.8511747371318729,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 86.8126,
+      "eval_samples_per_second": 15.839,
+      "eval_steps_per_second": 1.981,
+      "step": 1500
+    },
+    {
+      "entropy": 0.4021005939692259,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.8821248412132263,
+      "learning_rate": 0.00017710532822854468,
+      "loss": 0.3462103843688965,
+      "mean_token_accuracy": 0.889109355956316,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.4502199075596277,
+      "eval_loss": 0.566046416759491,
+      "eval_mean_token_accuracy": 0.8501714208098345,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 86.8336,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.981,
+      "step": 1520
+    },
+    {
+      "entropy": 0.4017397932708263,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8400952816009521,
+      "learning_rate": 0.0001754750908943189,
+      "loss": 0.34890995025634763,
+      "mean_token_accuracy": 0.8892098367214203,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.4614003023435903,
+      "eval_loss": 0.5617933869361877,
+      "eval_mean_token_accuracy": 0.8515863616106122,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 1540
+    },
+    {
+      "entropy": 0.4112051840871572,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 0.769478440284729,
+      "learning_rate": 0.0001738260540649939,
+      "loss": 0.34711437225341796,
+      "mean_token_accuracy": 0.8911717928946018,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4540443811998811,
+      "eval_loss": 0.5576469898223877,
+      "eval_mean_token_accuracy": 0.8512079674144124,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 86.5103,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 1560
+    },
+    {
+      "entropy": 0.41105241514742374,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.8468427062034607,
+      "learning_rate": 0.00017215871506758568,
+      "loss": 0.3433023452758789,
+      "mean_token_accuracy": 0.8898739732801915,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.4707539707075718,
+      "eval_loss": 0.5641466379165649,
+      "eval_mean_token_accuracy": 0.8495440957851188,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 86.609,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.986,
+      "step": 1580
+    },
+    {
+      "entropy": 0.41016379147768023,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.7482675313949585,
+      "learning_rate": 0.0001704735767487946,
+      "loss": 0.34550890922546384,
+      "mean_token_accuracy": 0.8893028847873211,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.46391099864660307,
+      "eval_loss": 0.5593640804290771,
+      "eval_mean_token_accuracy": 0.8510130581467651,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.3975,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 1600
+    },
+    {
+      "entropy": 0.33167599791135544,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9435692429542542,
+      "learning_rate": 0.00016877114732335337,
+      "loss": 0.2716026544570923,
+      "mean_token_accuracy": 0.9133149828666296,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.38499350005457567,
+      "eval_loss": 0.6298249363899231,
+      "eval_mean_token_accuracy": 0.8488117071778275,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.2933,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1620
+    },
+    {
+      "entropy": 0.3000166634097695,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.8080845475196838,
+      "learning_rate": 0.0001670519402207569,
+      "loss": 0.22617182731628419,
+      "mean_token_accuracy": 0.9253474645316601,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.370110988703578,
+      "eval_loss": 0.6338461637496948,
+      "eval_mean_token_accuracy": 0.8485634801692741,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 85.9508,
+      "eval_samples_per_second": 15.998,
+      "eval_steps_per_second": 2.001,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2986910421401262,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.7310900092124939,
+      "learning_rate": 0.0001653164739304185,
+      "loss": 0.22367463111877442,
+      "mean_token_accuracy": 0.9252275295555592,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.3944379702037157,
+      "eval_loss": 0.6109381914138794,
+      "eval_mean_token_accuracy": 0.849291454220927,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.6728,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.984,
+      "step": 1660
+    },
+    {
+      "entropy": 0.3095553796738386,
+      "epoch": 4.179327521793275,
+      "grad_norm": 0.7059140801429749,
+      "learning_rate": 0.0001635652718453007,
+      "loss": 0.23651680946350098,
+      "mean_token_accuracy": 0.9208931416273117,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.3910588648949945,
+      "eval_loss": 0.6104469299316406,
+      "eval_mean_token_accuracy": 0.8486883893262508,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7612,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.982,
+      "step": 1680
+    },
+    {
+      "entropy": 0.3001101028174162,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6787802577018738,
+      "learning_rate": 0.00016179886210406728,
+      "loss": 0.23130471706390382,
+      "mean_token_accuracy": 0.9233332790434361,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3794369170832079,
+      "eval_loss": 0.6182110905647278,
+      "eval_mean_token_accuracy": 0.8495433777570724,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 85.94,
+      "eval_samples_per_second": 16.0,
+      "eval_steps_per_second": 2.001,
+      "step": 1700
+    },
+    {
+      "entropy": 0.3031421799212694,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 0.9732038378715515,
+      "learning_rate": 0.0001600177774318036,
+      "loss": 0.2359529733657837,
+      "mean_token_accuracy": 0.9217648565769195,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3923123094231583,
+      "eval_loss": 0.6057384610176086,
+      "eval_mean_token_accuracy": 0.8508818288182103,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7647,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.29365369994193313,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7681498527526855,
+      "learning_rate": 0.0001582225549793541,
+      "loss": 0.2269371747970581,
+      "mean_token_accuracy": 0.9245341829955578,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.4011661055129628,
+      "eval_loss": 0.6144486665725708,
+      "eval_mean_token_accuracy": 0.8480324357054955,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 87.1306,
+      "eval_samples_per_second": 15.781,
+      "eval_steps_per_second": 1.974,
+      "step": 1740
+    },
+    {
+      "entropy": 0.29396994728595016,
+      "epoch": 4.378580323785803,
+      "grad_norm": 1.0001007318496704,
+      "learning_rate": 0.0001564137361613248,
+      "loss": 0.22777395248413085,
+      "mean_token_accuracy": 0.9262309700250626,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38518730195802314,
+      "eval_loss": 0.6202630400657654,
+      "eval_mean_token_accuracy": 0.8493869807137999,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6616,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.3096018506214023,
+      "epoch": 4.428393524283935,
+      "grad_norm": 1.0448365211486816,
+      "learning_rate": 0.00015459186649280024,
+      "loss": 0.23696351051330566,
+      "mean_token_accuracy": 0.9217322513461113,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.3946371126140273,
+      "eval_loss": 0.6079026460647583,
+      "eval_mean_token_accuracy": 0.8492515852978063,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6582,
+      "eval_samples_per_second": 15.867,
+      "eval_steps_per_second": 1.985,
+      "step": 1780
+    },
+    {
+      "entropy": 0.32619857545942066,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7210651636123657,
+      "learning_rate": 0.00015275749542482337,
+      "loss": 0.24651215076446534,
+      "mean_token_accuracy": 0.9177676141262054,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.3947690814560236,
+      "eval_loss": 0.6065912246704102,
+      "eval_mean_token_accuracy": 0.8502957744653835,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.5959,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.986,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3193941755220294,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.8281906843185425,
+      "learning_rate": 0.0001509111761786888,
+      "loss": 0.23936262130737304,
+      "mean_token_accuracy": 0.9201708927750587,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38704028864239537,
+      "eval_loss": 0.6006569266319275,
+      "eval_mean_token_accuracy": 0.8502406720505205,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.8059,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.981,
+      "step": 1820
+    },
+    {
+      "entropy": 0.3164879363030195,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.7892968654632568,
+      "learning_rate": 0.00014905346557909867,
+      "loss": 0.24541733264923096,
+      "mean_token_accuracy": 0.9175932116806507,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.38861122120951497,
+      "eval_loss": 0.6115967631340027,
+      "eval_mean_token_accuracy": 0.849471275196519,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.2946,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 1840
+    },
+    {
+      "entropy": 0.3051785985007882,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8109654188156128,
+      "learning_rate": 0.0001471849238862319,
+      "loss": 0.23433220386505127,
+      "mean_token_accuracy": 0.9206570319831371,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.37162452295076015,
+      "eval_loss": 0.6184061765670776,
+      "eval_mean_token_accuracy": 0.8501173268223918,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.6865,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.984,
+      "step": 1860
+    },
+    {
+      "entropy": 0.3168198253959417,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.9512342214584351,
+      "learning_rate": 0.0001453061146267775,
+      "loss": 0.23832404613494873,
+      "mean_token_accuracy": 0.9197044663131237,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3845940856912801,
+      "eval_loss": 0.606762707233429,
+      "eval_mean_token_accuracy": 0.8504838194957999,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.5175,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 1880
+    },
+    {
+      "entropy": 0.30791807882487776,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.8123113512992859,
+      "learning_rate": 0.00014341760442398248,
+      "loss": 0.2395785331726074,
+      "mean_token_accuracy": 0.918928150832653,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.39762327222283494,
+      "eval_loss": 0.5994202494621277,
+      "eval_mean_token_accuracy": 0.8509274201337681,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.2873,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.993,
+      "step": 1900
+    },
+    {
+      "entropy": 0.3021434534341097,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.731787383556366,
+      "learning_rate": 0.000141519962826766,
+      "loss": 0.23494718074798585,
+      "mean_token_accuracy": 0.9201403826475143,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.3827026732439219,
+      "eval_loss": 0.5995895862579346,
+      "eval_mean_token_accuracy": 0.851468373523202,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.3006,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 1920
+    },
+    {
+      "entropy": 0.31626159623265265,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8848487138748169,
+      "learning_rate": 0.00013961376213795132,
+      "loss": 0.2439030647277832,
+      "mean_token_accuracy": 0.9196575872600079,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.388698436839636,
+      "eval_loss": 0.6000174283981323,
+      "eval_mean_token_accuracy": 0.8518068187458571,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.8979,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.979,
+      "step": 1940
+    },
+    {
+      "entropy": 0.30520407035946845,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8532460927963257,
+      "learning_rate": 0.00013769957724166695,
+      "loss": 0.23458616733551024,
+      "mean_token_accuracy": 0.9221912942826748,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.38777847102908203,
+      "eval_loss": 0.6004981398582458,
+      "eval_mean_token_accuracy": 0.8516481768253238,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 87.0777,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.975,
+      "step": 1960
+    },
+    {
+      "entropy": 0.3226448342204094,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.6945561766624451,
+      "learning_rate": 0.0001357779854299694,
+      "loss": 0.24048397541046143,
+      "mean_token_accuracy": 0.9195300146937371,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.38581624263247777,
+      "eval_loss": 0.6029234528541565,
+      "eval_mean_token_accuracy": 0.8514213260523108,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 85.8729,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 1980
+    },
+    {
+      "entropy": 0.3051655298098922,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.7976452708244324,
+      "learning_rate": 0.00013384956622874001,
+      "loss": 0.23584742546081544,
+      "mean_token_accuracy": 0.9216851457953453,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.37913159246361533,
+      "eval_loss": 0.6057604551315308,
+      "eval_mean_token_accuracy": 0.8525801203971686,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.1145,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2000
+    },
+    {
+      "entropy": 0.27438195240803254,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.6729586124420166,
+      "learning_rate": 0.0001319149012229075,
+      "loss": 0.19775952100753785,
+      "mean_token_accuracy": 0.9339428559327737,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.3448961910813354,
+      "eval_loss": 0.6750120520591736,
+      "eval_mean_token_accuracy": 0.8487970232963562,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.1169,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 2020
+    },
+    {
+      "entropy": 0.21423916313797237,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.6934391856193542,
+      "learning_rate": 0.00012997457388105022,
+      "loss": 0.1439570426940918,
+      "mean_token_accuracy": 0.9528236843645572,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.3570949243771475,
+      "eval_loss": 0.6465504169464111,
+      "eval_mean_token_accuracy": 0.8490785547467166,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 85.9574,
+      "eval_samples_per_second": 15.996,
+      "eval_steps_per_second": 2.001,
+      "step": 2040
+    },
+    {
+      "entropy": 0.20838565267622472,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7286986112594604,
+      "learning_rate": 0.00012802916937942972,
+      "loss": 0.14467307329177856,
+      "mean_token_accuracy": 0.950994835793972,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.3452463157821533,
+      "eval_loss": 0.6705958843231201,
+      "eval_mean_token_accuracy": 0.8490352796953778,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 86.228,
+      "eval_samples_per_second": 15.946,
+      "eval_steps_per_second": 1.995,
+      "step": 2060
+    },
+    {
+      "entropy": 0.20453082229942082,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7515555620193481,
+      "learning_rate": 0.00012607927442550974,
+      "loss": 0.13732000589370727,
+      "mean_token_accuracy": 0.9537357829511166,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.32431264914745506,
+      "eval_loss": 0.6743043065071106,
+      "eval_mean_token_accuracy": 0.8504878629085629,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.5948,
+      "eval_samples_per_second": 15.879,
+      "eval_steps_per_second": 1.986,
+      "step": 2080
+    },
+    {
+      "entropy": 0.21812320686876774,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.9093465209007263,
+      "learning_rate": 0.0001241254770810132,
+      "loss": 0.14311420917510986,
+      "mean_token_accuracy": 0.9514068141579628,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.33086285835435225,
+      "eval_loss": 0.6676449179649353,
+      "eval_mean_token_accuracy": 0.8505287662495015,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 2100
+    },
+    {
+      "entropy": 0.2180163251236081,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6703007221221924,
+      "learning_rate": 0.00012216836658457075,
+      "loss": 0.14803968667984008,
+      "mean_token_accuracy": 0.9521303348243236,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.33162341708707255,
+      "eval_loss": 0.6658875942230225,
+      "eval_mean_token_accuracy": 0.8500757605530495,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.6296,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 2120
+    },
+    {
+      "entropy": 0.22511130161583423,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8078880906105042,
+      "learning_rate": 0.00012020853317401455,
+      "loss": 0.15228408575057983,
+      "mean_token_accuracy": 0.947144789993763,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3354601170434508,
+      "eval_loss": 0.6677829623222351,
+      "eval_mean_token_accuracy": 0.8482600024273229,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.4689,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.989,
+      "step": 2140
+    },
+    {
+      "entropy": 0.2244176059961319,
+      "epoch": 5.37359900373599,
+      "grad_norm": 0.9636075496673584,
+      "learning_rate": 0.00011824656790837037,
+      "loss": 0.15260883569717407,
+      "mean_token_accuracy": 0.9471467643976211,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.3381616926297199,
+      "eval_loss": 0.6694412231445312,
+      "eval_mean_token_accuracy": 0.8482369603805764,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.4147,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 2160
+    },
+    {
+      "entropy": 0.22982364390045404,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.775401771068573,
+      "learning_rate": 0.00011628306248960262,
+      "loss": 0.15140302181243898,
+      "mean_token_accuracy": 0.9492553934454918,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.3305150235808173,
+      "eval_loss": 0.6717822551727295,
+      "eval_mean_token_accuracy": 0.8489849723355715,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.4728,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.989,
+      "step": 2180
+    },
+    {
+      "entropy": 0.21448935717344284,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6575281023979187,
+      "learning_rate": 0.00011431860908416465,
+      "loss": 0.1452319622039795,
+      "mean_token_accuracy": 0.9505287684500218,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3488145174328671,
+      "eval_loss": 0.6612305641174316,
+      "eval_mean_token_accuracy": 0.8492907808963642,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6927,
+      "eval_samples_per_second": 15.861,
+      "eval_steps_per_second": 1.984,
+      "step": 2200
+    },
+    {
+      "entropy": 0.23091202937066554,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.8909686207771301,
+      "learning_rate": 0.00011235380014440985,
+      "loss": 0.15150139331817628,
+      "mean_token_accuracy": 0.9497875183820724,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.3268539015810157,
+      "eval_loss": 0.6667542457580566,
+      "eval_mean_token_accuracy": 0.8499062903398691,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.4644,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 2220
+    },
+    {
+      "entropy": 0.2227974807843566,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.6180275082588196,
+      "learning_rate": 0.0001103892282299158,
+      "loss": 0.1491069197654724,
+      "mean_token_accuracy": 0.9488144010305405,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3346347655494546,
+      "eval_loss": 0.6665948629379272,
+      "eval_mean_token_accuracy": 0.8499961893918903,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 87.0548,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.976,
+      "step": 2240
+    },
+    {
+      "entropy": 0.21368421968072654,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.6004369258880615,
+      "learning_rate": 0.00010842548582877651,
+      "loss": 0.14485255479812623,
+      "mean_token_accuracy": 0.9502056457102299,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.32753298804163933,
+      "eval_loss": 0.6680151224136353,
+      "eval_mean_token_accuracy": 0.8515451086121936,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.8524,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.98,
+      "step": 2260
+    },
+    {
+      "entropy": 0.2103635374456644,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.699174702167511,
+      "learning_rate": 0.00010646316517891613,
+      "loss": 0.14297772645950318,
+      "mean_token_accuracy": 0.9512537539005279,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.32966585959806,
+      "eval_loss": 0.675578773021698,
+      "eval_mean_token_accuracy": 0.8509623023659684,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.4518,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2280
+    },
+    {
+      "entropy": 0.22516900151968003,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6637354493141174,
+      "learning_rate": 0.00010450285808947797,
+      "loss": 0.15202572345733642,
+      "mean_token_accuracy": 0.9482452072203159,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3369456917740578,
+      "eval_loss": 0.6697061061859131,
+      "eval_mean_token_accuracy": 0.848603522361711,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.6371,
+      "eval_samples_per_second": 15.871,
+      "eval_steps_per_second": 1.985,
+      "step": 2300
+    },
+    {
+      "entropy": 0.21841065725311637,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7940268516540527,
+      "learning_rate": 0.00010254515576234297,
+      "loss": 0.14710167646408082,
+      "mean_token_accuracy": 0.9493498183786869,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3237486180177955,
+      "eval_loss": 0.6779657602310181,
+      "eval_mean_token_accuracy": 0.8500715313955794,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.1826,
+      "eval_samples_per_second": 15.954,
+      "eval_steps_per_second": 1.996,
+      "step": 2320
+    },
+    {
+      "entropy": 0.22146204356104135,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.9234833121299744,
+      "learning_rate": 0.00010059064861383132,
+      "loss": 0.14720046520233154,
+      "mean_token_accuracy": 0.9493872679769992,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.32365207564692167,
+      "eval_loss": 0.6704005002975464,
+      "eval_mean_token_accuracy": 0.8507985760306203,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.5494,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.987,
+      "step": 2340
+    },
+    {
+      "entropy": 0.20934011954814197,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.5547503232955933,
+      "learning_rate": 9.863992609664084e-05,
+      "loss": 0.1464867353439331,
+      "mean_token_accuracy": 0.9503875687718392,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.3330401429083458,
+      "eval_loss": 0.6659091114997864,
+      "eval_mean_token_accuracy": 0.8504848906467127,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.5855,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 2360
+    },
+    {
+      "entropy": 0.21678635366261007,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.570612907409668,
+      "learning_rate": 9.669357652207635e-05,
+      "loss": 0.14821385145187377,
+      "mean_token_accuracy": 0.9503940217196941,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3322022325077722,
+      "eval_loss": 0.6722489595413208,
+      "eval_mean_token_accuracy": 0.8489979422369669,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.2986,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 2380
+    },
+    {
+      "entropy": 0.20932920072227718,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7879557609558105,
+      "learning_rate": 9.475218688262262e-05,
+      "loss": 0.14675841331481934,
+      "mean_token_accuracy": 0.9507176131010056,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.3199348642902319,
+      "eval_loss": 0.6698136329650879,
+      "eval_mean_token_accuracy": 0.8514142130003419,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 85.8995,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 2400
+    },
+    {
+      "entropy": 0.21032143919131693,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.5823076367378235,
+      "learning_rate": 9.281634267491569e-05,
+      "loss": 0.13322267532348633,
+      "mean_token_accuracy": 0.9550939072401096,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.30206270117399303,
+      "eval_loss": 0.7093168497085571,
+      "eval_mean_token_accuracy": 0.8500627639681794,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 85.8128,
+      "eval_samples_per_second": 16.023,
+      "eval_steps_per_second": 2.004,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1534628233872354,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.710133969783783,
+      "learning_rate": 9.088662772316508e-05,
+      "loss": 0.09078952670097351,
+      "mean_token_accuracy": 0.9678447999060154,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.28208133101809857,
+      "eval_loss": 0.7636417150497437,
+      "eval_mean_token_accuracy": 0.8494061477655588,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9724,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 2.001,
+      "step": 2440
+    },
+    {
+      "entropy": 0.16151462448760867,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.5793812274932861,
+      "learning_rate": 8.896362400308028e-05,
+      "loss": 0.09545697569847107,
+      "mean_token_accuracy": 0.9671573750674725,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.2833245605403601,
+      "eval_loss": 0.7402405738830566,
+      "eval_mean_token_accuracy": 0.8503523728875226,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 85.5461,
+      "eval_samples_per_second": 16.073,
+      "eval_steps_per_second": 2.011,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15203177919611335,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.4251123368740082,
+      "learning_rate": 8.704791146635483e-05,
+      "loss": 0.09420782327651978,
+      "mean_token_accuracy": 0.9677386932075024,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.28572545962971313,
+      "eval_loss": 0.735416829586029,
+      "eval_mean_token_accuracy": 0.8487084663884584,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 85.4801,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.012,
+      "step": 2480
+    },
+    {
+      "entropy": 0.15587336672469973,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.4357028007507324,
+      "learning_rate": 8.514006786576085e-05,
+      "loss": 0.09429320096969604,
+      "mean_token_accuracy": 0.9682952925562859,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.2859006894882335,
+      "eval_loss": 0.7347224950790405,
+      "eval_mean_token_accuracy": 0.8501905518215757,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 85.7578,
+      "eval_samples_per_second": 16.034,
+      "eval_steps_per_second": 2.006,
+      "step": 2500
+    },
+    {
+      "entropy": 0.15765639198943973,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.47331130504608154,
+      "learning_rate": 8.324066858090663e-05,
+      "loss": 0.09571113586425781,
+      "mean_token_accuracy": 0.9683481335639954,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.29231513846059176,
+      "eval_loss": 0.7392512559890747,
+      "eval_mean_token_accuracy": 0.8486633983462356,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 85.7846,
+      "eval_samples_per_second": 16.029,
+      "eval_steps_per_second": 2.005,
+      "step": 2520
+    },
+    {
+      "entropy": 0.16176860257983208,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.6142018437385559,
+      "learning_rate": 8.135028644470992e-05,
+      "loss": 0.09486144185066223,
+      "mean_token_accuracy": 0.9682316601276397,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.2851274753379267,
+      "eval_loss": 0.7520816922187805,
+      "eval_mean_token_accuracy": 0.8501113649717597,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 85.9425,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 2540
+    },
+    {
+      "entropy": 0.15404034564271568,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6051287651062012,
+      "learning_rate": 7.946949157063964e-05,
+      "loss": 0.09280472993850708,
+      "mean_token_accuracy": 0.9684635892510414,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28802703563557114,
+      "eval_loss": 0.7439162135124207,
+      "eval_mean_token_accuracy": 0.8499851770872293,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.0703,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.998,
+      "step": 2560
+    },
+    {
+      "entropy": 0.15343588953837753,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.4823743402957916,
+      "learning_rate": 7.759885118077701e-05,
+      "loss": 0.09000591039657593,
+      "mean_token_accuracy": 0.9699928767979145,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2795634938533916,
+      "eval_loss": 0.7647850513458252,
+      "eval_mean_token_accuracy": 0.8503464397995971,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 85.8886,
+      "eval_samples_per_second": 16.009,
+      "eval_steps_per_second": 2.003,
+      "step": 2580
+    },
+    {
+      "entropy": 0.15740026142448188,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.5082696676254272,
+      "learning_rate": 7.57389294347494e-05,
+      "loss": 0.09191849827766418,
+      "mean_token_accuracy": 0.9692809768021107,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2913342311458532,
+      "eval_loss": 0.7518694996833801,
+      "eval_mean_token_accuracy": 0.8483168302580367,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 85.5761,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.01,
+      "step": 2600
+    },
+    {
+      "entropy": 0.15922069251537324,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.3995199501514435,
+      "learning_rate": 7.389028725958736e-05,
+      "loss": 0.09584367871284485,
+      "mean_token_accuracy": 0.9685114495456218,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.2863075934177221,
+      "eval_loss": 0.7539381384849548,
+      "eval_mean_token_accuracy": 0.8507507083027862,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.112,
+      "eval_samples_per_second": 15.968,
+      "eval_steps_per_second": 1.997,
+      "step": 2620
+    },
+    {
+      "entropy": 0.16197575423866512,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.534295380115509,
+      "learning_rate": 7.205348218055678e-05,
+      "loss": 0.0961170256137848,
+      "mean_token_accuracy": 0.9674530044198036,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.29021967315050057,
+      "eval_loss": 0.751198947429657,
+      "eval_mean_token_accuracy": 0.8509796344956686,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 85.7958,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 2640
+    },
+    {
+      "entropy": 0.15261746793985367,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.5391237139701843,
+      "learning_rate": 7.022906815301673e-05,
+      "loss": 0.0926026999950409,
+      "mean_token_accuracy": 0.9695659473538398,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.29128045216202736,
+      "eval_loss": 0.7450292110443115,
+      "eval_mean_token_accuracy": 0.8498771443616512,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.3963,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 2660
+    },
+    {
+      "entropy": 0.16164180357009172,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.5767946243286133,
+      "learning_rate": 6.841759539535419e-05,
+      "loss": 0.09291981458663941,
+      "mean_token_accuracy": 0.9687136687338352,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2897637223088464,
+      "eval_loss": 0.7503410577774048,
+      "eval_mean_token_accuracy": 0.8503315164599308,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.0653,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.998,
+      "step": 2680
+    },
+    {
+      "entropy": 0.17062523355707526,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.4974168539047241,
+      "learning_rate": 6.661961022304563e-05,
+      "loss": 0.09713236689567566,
+      "mean_token_accuracy": 0.9661971725523472,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2765843396963075,
+      "eval_loss": 0.7604002356529236,
+      "eval_mean_token_accuracy": 0.8521115277395692,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.1676,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 2700
+    },
+    {
+      "entropy": 0.17544092936441302,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.5523537993431091,
+      "learning_rate": 6.483565488389582e-05,
+      "loss": 0.09709116220474243,
+      "mean_token_accuracy": 0.9650957375764847,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.27939334659035814,
+      "eval_loss": 0.7534670829772949,
+      "eval_mean_token_accuracy": 0.851230604010959,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.2913,
+      "eval_samples_per_second": 15.934,
+      "eval_steps_per_second": 1.993,
+      "step": 2720
+    },
+    {
+      "entropy": 0.15991022661328316,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.478551983833313,
+      "learning_rate": 6.306626739450311e-05,
+      "loss": 0.09564646482467651,
+      "mean_token_accuracy": 0.9679084822535515,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.27878295491601146,
+      "eval_loss": 0.7519959211349487,
+      "eval_mean_token_accuracy": 0.8510654949864676,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 85.9109,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16824879590421915,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.6681098937988281,
+      "learning_rate": 6.131198137800108e-05,
+      "loss": 0.0962279736995697,
+      "mean_token_accuracy": 0.966830012947321,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.2834690434121808,
+      "eval_loss": 0.751922070980072,
+      "eval_mean_token_accuracy": 0.8521237577809844,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.3618,
+      "eval_samples_per_second": 15.921,
+      "eval_steps_per_second": 1.992,
+      "step": 2760
+    },
+    {
+      "entropy": 0.1518704930320382,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5201290249824524,
+      "learning_rate": 5.957332590312513e-05,
+      "loss": 0.09010660648345947,
+      "mean_token_accuracy": 0.9693463340401649,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.28485129617674404,
+      "eval_loss": 0.7452457547187805,
+      "eval_mean_token_accuracy": 0.8512036796919135,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.4524,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.99,
+      "step": 2780
+    },
+    {
+      "entropy": 0.15512610590085388,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.42802050709724426,
+      "learning_rate": 5.785082532465233e-05,
+      "loss": 0.09320432543754578,
+      "mean_token_accuracy": 0.9686134628951549,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.27554594526110693,
+      "eval_loss": 0.7644653916358948,
+      "eval_mean_token_accuracy": 0.8513738523389018,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 85.7609,
+      "eval_samples_per_second": 16.033,
+      "eval_steps_per_second": 2.006,
+      "step": 2800
+    },
+    {
+      "entropy": 0.17524497526196334,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.3330269157886505,
+      "learning_rate": 5.6144999125263545e-05,
+      "loss": 0.0895616888999939,
+      "mean_token_accuracy": 0.9682766932707566,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.2735865569218647,
+      "eval_loss": 0.768479585647583,
+      "eval_mean_token_accuracy": 0.8503459383581959,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 85.7162,
+      "eval_samples_per_second": 16.041,
+      "eval_steps_per_second": 2.007,
+      "step": 2820
+    },
+    {
+      "entropy": 0.1403565663844347,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.35613498091697693,
+      "learning_rate": 5.4456361758874235e-05,
+      "loss": 0.07551313638687134,
+      "mean_token_accuracy": 0.9739301167428493,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.25941789089593775,
+      "eval_loss": 0.8209511637687683,
+      "eval_mean_token_accuracy": 0.8505055855873019,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.0943,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 2840
+    },
+    {
+      "entropy": 0.13500106502324344,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.34418627619743347,
+      "learning_rate": 5.2785422495482234e-05,
+      "loss": 0.07293946146965027,
+      "mean_token_accuracy": 0.9747208289802074,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.26482899772912954,
+      "eval_loss": 0.798904538154602,
+      "eval_mean_token_accuracy": 0.8511530233677044,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 85.7915,
+      "eval_samples_per_second": 16.027,
+      "eval_steps_per_second": 2.005,
+      "step": 2860
+    },
+    {
+      "entropy": 0.13127502552233636,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.4638441503047943,
+      "learning_rate": 5.113268526757892e-05,
+      "loss": 0.07121461629867554,
+      "mean_token_accuracy": 0.9756786689162255,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2645381211714689,
+      "eval_loss": 0.809101939201355,
+      "eval_mean_token_accuracy": 0.8514727898115335,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.84,
+      "eval_samples_per_second": 16.018,
+      "eval_steps_per_second": 2.004,
+      "step": 2880
+    },
+    {
+      "entropy": 0.1331390908919275,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.40206775069236755,
+      "learning_rate": 4.949864851817007e-05,
+      "loss": 0.07188264131546021,
+      "mean_token_accuracy": 0.9755406074225903,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.26244733283339544,
+      "eval_loss": 0.8143188953399658,
+      "eval_mean_token_accuracy": 0.8514358189909957,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.8546,
+      "eval_samples_per_second": 16.015,
+      "eval_steps_per_second": 2.003,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13647331558167936,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.26405787467956543,
+      "learning_rate": 4.788380505045224e-05,
+      "loss": 0.07412450313568116,
+      "mean_token_accuracy": 0.9744274213910102,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2626111418182074,
+      "eval_loss": 0.8111525177955627,
+      "eval_mean_token_accuracy": 0.8512121695418691,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 85.9626,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 2920
+    },
+    {
+      "entropy": 0.12644002586603165,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.27514681220054626,
+      "learning_rate": 4.6288641879189884e-05,
+      "loss": 0.06807711124420165,
+      "mean_token_accuracy": 0.9760703906416893,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.26096762734097106,
+      "eval_loss": 0.8184595108032227,
+      "eval_mean_token_accuracy": 0.8501476153384807,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 85.9521,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13920630998909472,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.23584705591201782,
+      "learning_rate": 4.4713640083838604e-05,
+      "loss": 0.07301607131958007,
+      "mean_token_accuracy": 0.9745715200901032,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.2612536767021168,
+      "eval_loss": 0.8116245269775391,
+      "eval_mean_token_accuracy": 0.8510967350976412,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 85.6373,
+      "eval_samples_per_second": 16.056,
+      "eval_steps_per_second": 2.008,
+      "step": 2960
+    },
+    {
+      "entropy": 0.1349492883309722,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.24552719295024872,
+      "learning_rate": 4.315927466345791e-05,
+      "loss": 0.07397544980049134,
+      "mean_token_accuracy": 0.9745095103979111,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.25796533306670744,
+      "eval_loss": 0.8266491293907166,
+      "eval_mean_token_accuracy": 0.8511653857868772,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 85.7462,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 2980
+    },
+    {
+      "entropy": 0.14479175000451505,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.2846072018146515,
+      "learning_rate": 4.162601439345814e-05,
+      "loss": 0.07544798254966736,
+      "mean_token_accuracy": 0.9727819666266442,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.2584348309698493,
+      "eval_loss": 0.8253348469734192,
+      "eval_mean_token_accuracy": 0.8511569815319638,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.2984,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 3000
+    },
+    {
+      "entropy": 0.14114196319133043,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.407966285943985,
+      "learning_rate": 4.011432168422419e-05,
+      "loss": 0.0736398994922638,
+      "mean_token_accuracy": 0.9741654269397259,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.25232676260693127,
+      "eval_loss": 0.8296052813529968,
+      "eval_mean_token_accuracy": 0.8523298349491385,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 85.8445,
+      "eval_samples_per_second": 16.017,
+      "eval_steps_per_second": 2.004,
+      "step": 3020
+    },
+    {
+      "entropy": 0.1353456223383546,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.2712634801864624,
+      "learning_rate": 3.8624652441658684e-05,
+      "loss": 0.07362412214279175,
+      "mean_token_accuracy": 0.9747945807874203,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.2579477243991785,
+      "eval_loss": 0.8274564743041992,
+      "eval_mean_token_accuracy": 0.8517705212498821,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 85.8222,
+      "eval_samples_per_second": 16.022,
+      "eval_steps_per_second": 2.004,
+      "step": 3040
+    },
+    {
+      "entropy": 0.1296080862637609,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.2371893972158432,
+      "learning_rate": 3.715745592968707e-05,
+      "loss": 0.06971035599708557,
+      "mean_token_accuracy": 0.9759043246507645,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.25391967083479083,
+      "eval_loss": 0.8197130560874939,
+      "eval_mean_token_accuracy": 0.8522267875283264,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 85.8796,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1311203008517623,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.22499267756938934,
+      "learning_rate": 3.5713174634765967e-05,
+      "loss": 0.07176244258880615,
+      "mean_token_accuracy": 0.9754939571022987,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.2526215241225653,
+      "eval_loss": 0.8265154361724854,
+      "eval_mean_token_accuracy": 0.8519952584837758,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 85.8746,
+      "eval_samples_per_second": 16.012,
+      "eval_steps_per_second": 2.003,
+      "step": 3080
+    },
+    {
+      "entropy": 0.13420484317466616,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.2398064285516739,
+      "learning_rate": 3.4292244132434866e-05,
+      "loss": 0.07559212446212768,
+      "mean_token_accuracy": 0.9743142127990723,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.2484562756537005,
+      "eval_loss": 0.8312150239944458,
+      "eval_mean_token_accuracy": 0.8528405119513356,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 85.7896,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11965563679113984,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.3408384919166565,
+      "learning_rate": 3.2895092955952746e-05,
+      "loss": 0.0661750853061676,
+      "mean_token_accuracy": 0.9776600524783134,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2522421533804993,
+      "eval_loss": 0.8327009677886963,
+      "eval_mean_token_accuracy": 0.8524222023958383,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.1769,
+      "eval_samples_per_second": 15.956,
+      "eval_steps_per_second": 1.996,
+      "step": 3120
+    },
+    {
+      "entropy": 0.13388084415346385,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.35418516397476196,
+      "learning_rate": 3.152214246705829e-05,
+      "loss": 0.07149899601936341,
+      "mean_token_accuracy": 0.9747635535895824,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.25038352296795957,
+      "eval_loss": 0.836457371711731,
+      "eval_mean_token_accuracy": 0.8526130665180295,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 85.6099,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.009,
+      "step": 3140
+    },
+    {
+      "entropy": 0.13530643959529698,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.38060539960861206,
+      "learning_rate": 3.017380672889291e-05,
+      "loss": 0.07116585969924927,
+      "mean_token_accuracy": 0.973284512758255,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.255092611319797,
+      "eval_loss": 0.8314701914787292,
+      "eval_mean_token_accuracy": 0.8520913099826768,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 85.877,
+      "eval_samples_per_second": 16.011,
+      "eval_steps_per_second": 2.003,
+      "step": 3160
+    },
+    {
+      "entropy": 0.13383390177041293,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.3827536106109619,
+      "learning_rate": 2.8850492381124808e-05,
+      "loss": 0.07305437326431274,
+      "mean_token_accuracy": 0.9750778004527092,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.25416371157003004,
+      "eval_loss": 0.8206402063369751,
+      "eval_mean_token_accuracy": 0.852779901651449,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.1015,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3180
+    },
+    {
+      "entropy": 0.1395680439658463,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.3809775412082672,
+      "learning_rate": 2.7552598517312256e-05,
+      "loss": 0.07236042022705078,
+      "mean_token_accuracy": 0.9731538116931915,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.25528111975899964,
+      "eval_loss": 0.8230037093162537,
+      "eval_mean_token_accuracy": 0.8526452603035195,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 85.7895,
+      "eval_samples_per_second": 16.028,
+      "eval_steps_per_second": 2.005,
+      "step": 3200
+    },
+    {
+      "entropy": 0.12193699864049752,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11854425817728043,
+      "learning_rate": 2.6280516564542205e-05,
+      "loss": 0.06617764234542847,
+      "mean_token_accuracy": 0.977179691577569,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.25100527677771656,
+      "eval_loss": 0.8393343687057495,
+      "eval_mean_token_accuracy": 0.8522553132023922,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 85.8837,
+      "eval_samples_per_second": 16.01,
+      "eval_steps_per_second": 2.003,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12788885813206435,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.16094881296157837,
+      "learning_rate": 2.50346301653812e-05,
+      "loss": 0.06274186968803405,
+      "mean_token_accuracy": 0.9766994707286358,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.24409458682287571,
+      "eval_loss": 0.874617338180542,
+      "eval_mean_token_accuracy": 0.8521041180505309,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.2656,
+      "eval_samples_per_second": 15.939,
+      "eval_steps_per_second": 1.994,
+      "step": 3240
+    },
+    {
+      "entropy": 0.12464155335910618,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.16893954575061798,
+      "learning_rate": 2.381531506217437e-05,
+      "loss": 0.06093020439147949,
+      "mean_token_accuracy": 0.9776258453726768,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.24396493017326953,
+      "eval_loss": 0.891161322593689,
+      "eval_mean_token_accuracy": 0.8515338024427724,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 85.9061,
+      "eval_samples_per_second": 16.006,
+      "eval_steps_per_second": 2.002,
+      "step": 3260
+    },
+    {
+      "entropy": 0.12345920228399336,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.14332273602485657,
+      "learning_rate": 2.262293898372616e-05,
+      "loss": 0.061289966106414795,
+      "mean_token_accuracy": 0.9762230902910233,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.2481445314059424,
+      "eval_loss": 0.8922848105430603,
+      "eval_mean_token_accuracy": 0.8514944855556932,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 85.5201,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3280
+    },
+    {
+      "entropy": 0.12298110066913068,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.21848882734775543,
+      "learning_rate": 2.1457861534398633e-05,
+      "loss": 0.05986443758010864,
+      "mean_token_accuracy": 0.9786281704902648,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.24329388124305149,
+      "eval_loss": 0.9021085500717163,
+      "eval_mean_token_accuracy": 0.8521762625422589,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 85.955,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3300
+    },
+    {
+      "entropy": 0.13265180448070168,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.18067947030067444,
+      "learning_rate": 2.0320434085659692e-05,
+      "loss": 0.06724961400032044,
+      "mean_token_accuracy": 0.975098168104887,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.2433211464694766,
+      "eval_loss": 0.9094350337982178,
+      "eval_mean_token_accuracy": 0.8520150094531304,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 85.7989,
+      "eval_samples_per_second": 16.026,
+      "eval_steps_per_second": 2.005,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11949320202693343,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17020289599895477,
+      "learning_rate": 1.9210999670114196e-05,
+      "loss": 0.0634455680847168,
+      "mean_token_accuracy": 0.9771294385194779,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.24345201219237128,
+      "eval_loss": 0.9021793603897095,
+      "eval_mean_token_accuracy": 0.8520745697409607,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.0883,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 3340
+    },
+    {
+      "entropy": 0.12065747743472457,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.16789060831069946,
+      "learning_rate": 1.8129892878049886e-05,
+      "loss": 0.061494195461273195,
+      "mean_token_accuracy": 0.9779584899544715,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.24093415042342142,
+      "eval_loss": 0.9006755352020264,
+      "eval_mean_token_accuracy": 0.852174230786257,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 85.9011,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3360
+    },
+    {
+      "entropy": 0.13125578537583352,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1662452220916748,
+      "learning_rate": 1.70774397565298e-05,
+      "loss": 0.06685600876808166,
+      "mean_token_accuracy": 0.9744067586958408,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.24062153688350388,
+      "eval_loss": 0.9078915119171143,
+      "eval_mean_token_accuracy": 0.8523201647886011,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0201,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11986117120832204,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.19571948051452637,
+      "learning_rate": 1.6053957711060606e-05,
+      "loss": 0.06411095261573792,
+      "mean_token_accuracy": 0.9770627245306969,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.24352820347561394,
+      "eval_loss": 0.9058943390846252,
+      "eval_mean_token_accuracy": 0.8519382792156797,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 85.966,
+      "eval_samples_per_second": 15.995,
+      "eval_steps_per_second": 2.001,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12857897616922856,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.2609264850616455,
+      "learning_rate": 1.5059755409867613e-05,
+      "loss": 0.06473397612571716,
+      "mean_token_accuracy": 0.9764650195837021,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.24032609000108962,
+      "eval_loss": 0.9077776074409485,
+      "eval_mean_token_accuracy": 0.8517829197090726,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6059,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.12348870886489749,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.19537609815597534,
+      "learning_rate": 1.409513269080473e-05,
+      "loss": 0.06481348872184753,
+      "mean_token_accuracy": 0.9769559659063816,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.2404322574824788,
+      "eval_loss": 0.9057720899581909,
+      "eval_mean_token_accuracy": 0.8521037981953732,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.166,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.996,
+      "step": 3440
+    },
+    {
+      "entropy": 0.12040232736617326,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.3310582935810089,
+      "learning_rate": 1.3160380470927183e-05,
+      "loss": 0.06468871235847473,
+      "mean_token_accuracy": 0.9768650442361831,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.24118655876711356,
+      "eval_loss": 0.9028318524360657,
+      "eval_mean_token_accuracy": 0.8521025340224422,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 85.3668,
+      "eval_samples_per_second": 16.107,
+      "eval_steps_per_second": 2.015,
+      "step": 3460
+    },
+    {
+      "entropy": 0.12328103906475008,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.12836167216300964,
+      "learning_rate": 1.2255780658755122e-05,
+      "loss": 0.06229386329650879,
+      "mean_token_accuracy": 0.9763277888298034,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.24194598145956217,
+      "eval_loss": 0.9009329080581665,
+      "eval_mean_token_accuracy": 0.8521693289973015,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 85.9415,
+      "eval_samples_per_second": 15.999,
+      "eval_steps_per_second": 2.001,
+      "step": 3480
+    },
+    {
+      "entropy": 0.11321646976284683,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.15656894445419312,
+      "learning_rate": 1.1381606069253786e-05,
+      "loss": 0.05908188819885254,
+      "mean_token_accuracy": 0.9779504477977753,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.24121542517528977,
+      "eval_loss": 0.9016091823577881,
+      "eval_mean_token_accuracy": 0.8521790667328724,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 85.7465,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3500
+    },
+    {
+      "entropy": 0.12657046681270004,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.22597502171993256,
+      "learning_rate": 1.053812034155629e-05,
+      "loss": 0.06244581937789917,
+      "mean_token_accuracy": 0.976795207709074,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.23877542708502258,
+      "eval_loss": 0.9069110155105591,
+      "eval_mean_token_accuracy": 0.8522880177858264,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 85.7792,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3520
+    },
+    {
+      "entropy": 0.11422101808711886,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.21139323711395264,
+      "learning_rate": 9.725577859453502e-06,
+      "loss": 0.05988085865974426,
+      "mean_token_accuracy": 0.9779510758817196,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.2393161087881687,
+      "eval_loss": 0.9033801555633545,
+      "eval_mean_token_accuracy": 0.8522614209457885,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 85.5594,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 3540
+    },
+    {
+      "entropy": 0.13810604228638113,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.24571993947029114,
+      "learning_rate": 8.944223674675537e-06,
+      "loss": 0.07036117911338806,
+      "mean_token_accuracy": 0.9734892748296261,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.23998506947658782,
+      "eval_loss": 0.9009848833084106,
+      "eval_mean_token_accuracy": 0.8521793619837872,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0974,
+      "eval_samples_per_second": 15.97,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.14193559321574867,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.17304112017154694,
+      "learning_rate": 8.194293432987386e-06,
+      "loss": 0.07077967524528503,
+      "mean_token_accuracy": 0.973305893689394,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.23883876689644748,
+      "eval_loss": 0.9015622138977051,
+      "eval_mean_token_accuracy": 0.8524864378363587,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.0089,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 2.0,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11878943420015275,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.19075658917427063,
+      "learning_rate": 7.476013303121426e-06,
+      "loss": 0.060806107521057126,
+      "mean_token_accuracy": 0.9776863709092141,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.23726526309931,
+      "eval_loss": 0.9060276746749878,
+      "eval_mean_token_accuracy": 0.8524192058762838,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 85.7252,
+      "eval_samples_per_second": 16.04,
+      "eval_steps_per_second": 2.006,
+      "step": 3600
+    },
+    {
+      "entropy": 0.1255835090787747,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11608047038316727,
+      "learning_rate": 6.78959990856799e-06,
+      "loss": 0.06319612860679627,
+      "mean_token_accuracy": 0.9766685519462976,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2373251837006835,
+      "eval_loss": 0.9045722484588623,
+      "eval_mean_token_accuracy": 0.8525558363559634,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 85.7435,
+      "eval_samples_per_second": 16.036,
+      "eval_steps_per_second": 2.006,
+      "step": 3620
+    },
+    {
+      "entropy": 0.12587778437882663,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1564614623785019,
+      "learning_rate": 6.135260262244683e-06,
+      "loss": 0.0630365788936615,
+      "mean_token_accuracy": 0.9777486085891723,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.23559923721260803,
+      "eval_loss": 0.9120694398880005,
+      "eval_mean_token_accuracy": 0.8525292966947999,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 85.8018,
+      "eval_samples_per_second": 16.025,
+      "eval_steps_per_second": 2.005,
+      "step": 3640
+    },
+    {
+      "entropy": 0.12535365503281354,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.1404249221086502,
+      "learning_rate": 5.513191704064086e-06,
+      "loss": 0.060502654314041136,
+      "mean_token_accuracy": 0.9770058333873749,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.23525122691725575,
+      "eval_loss": 0.9182237982749939,
+      "eval_mean_token_accuracy": 0.8524098333924316,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 85.9872,
+      "eval_samples_per_second": 15.991,
+      "eval_steps_per_second": 2.0,
+      "step": 3660
+    },
+    {
+      "entropy": 0.11330419047735632,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15513843297958374,
+      "learning_rate": 4.92358184141876e-06,
+      "loss": 0.05822383761405945,
+      "mean_token_accuracy": 0.9793384782969952,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.2353947116711805,
+      "eval_loss": 0.9229223132133484,
+      "eval_mean_token_accuracy": 0.852500357253607,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.0805,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.998,
+      "step": 3680
+    },
+    {
+      "entropy": 0.11830627010203898,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.1730550080537796,
+      "learning_rate": 4.366608492601456e-06,
+      "loss": 0.05860854983329773,
+      "mean_token_accuracy": 0.9779663667082786,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.23567205719476522,
+      "eval_loss": 0.9269373416900635,
+      "eval_mean_token_accuracy": 0.8523658004611038,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 85.9809,
+      "eval_samples_per_second": 15.992,
+      "eval_steps_per_second": 2.0,
+      "step": 3700
+    },
+    {
+      "entropy": 0.1180900705512613,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.20909737050533295,
+      "learning_rate": 3.842439633177387e-06,
+      "loss": 0.059438884258270264,
+      "mean_token_accuracy": 0.9786856278777123,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.23602714493524196,
+      "eval_loss": 0.9293538928031921,
+      "eval_mean_token_accuracy": 0.8523273440294488,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 85.2118,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.019,
+      "step": 3720
+    },
+    {
+      "entropy": 0.13156692241318524,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12535709142684937,
+      "learning_rate": 3.3512333453253305e-06,
+      "loss": 0.06337688565254211,
+      "mean_token_accuracy": 0.9756712593138218,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.23534389351343,
+      "eval_loss": 0.932999312877655,
+      "eval_mean_token_accuracy": 0.8523333925147389,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 85.953,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3740
+    },
+    {
+      "entropy": 0.12327516414225101,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.16341575980186462,
+      "learning_rate": 2.8931377701619306e-06,
+      "loss": 0.05869622826576233,
+      "mean_token_accuracy": 0.9784224212169648,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.23493653622477553,
+      "eval_loss": 0.9358566999435425,
+      "eval_mean_token_accuracy": 0.8522722783476807,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 85.2538,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.018,
+      "step": 3760
+    },
+    {
+      "entropy": 0.1134357453789562,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.23999616503715515,
+      "learning_rate": 2.4682910630645723e-06,
+      "loss": 0.057540220022201535,
+      "mean_token_accuracy": 0.9798057802021504,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.23470525634150172,
+      "eval_loss": 0.937556266784668,
+      "eval_mean_token_accuracy": 0.8523351706044618,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 85.7764,
+      "eval_samples_per_second": 16.03,
+      "eval_steps_per_second": 2.005,
+      "step": 3780
+    },
+    {
+      "entropy": 0.1075570048764348,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.15417765080928802,
+      "learning_rate": 2.0768213520055406e-06,
+      "loss": 0.05581026673316956,
+      "mean_token_accuracy": 0.9797527104616165,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.2347462713545145,
+      "eval_loss": 0.9383137226104736,
+      "eval_mean_token_accuracy": 0.8522575324357942,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 85.8985,
+      "eval_samples_per_second": 16.007,
+      "eval_steps_per_second": 2.002,
+      "step": 3800
+    },
+    {
+      "entropy": 0.12609313456341625,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.22041426599025726,
+      "learning_rate": 1.7188466989102739e-06,
+      "loss": 0.06379218697547913,
+      "mean_token_accuracy": 0.9763593293726445,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.23459658849724505,
+      "eval_loss": 0.9393337965011597,
+      "eval_mean_token_accuracy": 0.8523633532052817,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.9348,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.002,
+      "step": 3820
+    },
+    {
+      "entropy": 0.12149709439836442,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.16608643531799316,
+      "learning_rate": 1.3944750640516357e-06,
+      "loss": 0.06341606974601746,
+      "mean_token_accuracy": 0.9771503552794456,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.2347291322468325,
+      "eval_loss": 0.9399036765098572,
+      "eval_mean_token_accuracy": 0.8523768914300341,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.1305,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.997,
+      "step": 3840
+    },
+    {
+      "entropy": 0.11724846777506173,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13613300025463104,
+      "learning_rate": 1.103804273490497e-06,
+      "loss": 0.05994418263435364,
+      "mean_token_accuracy": 0.9778759330511093,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.23495923337894817,
+      "eval_loss": 0.9400841593742371,
+      "eval_mean_token_accuracy": 0.8523780471363733,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.0379,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.999,
+      "step": 3860
+    },
+    {
+      "entropy": 0.12054574331268668,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11884245276451111,
+      "learning_rate": 8.469219895727582e-07,
+      "loss": 0.05917409062385559,
+      "mean_token_accuracy": 0.9771256923675538,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.23499552723626757,
+      "eval_loss": 0.9404177665710449,
+      "eval_mean_token_accuracy": 0.8523571678372317,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.0174,
+      "eval_samples_per_second": 15.985,
+      "eval_steps_per_second": 2.0,
+      "step": 3880
+    },
+    {
+      "entropy": 0.12163264504633844,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.18393972516059875,
+      "learning_rate": 6.239056844915407e-07,
+      "loss": 0.059613007307052615,
+      "mean_token_accuracy": 0.9776720523834228,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.23491846319547918,
+      "eval_loss": 0.9405836462974548,
+      "eval_mean_token_accuracy": 0.8523543633000795,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 85.9519,
+      "eval_samples_per_second": 15.997,
+      "eval_steps_per_second": 2.001,
+      "step": 3900
+    },
+    {
+      "entropy": 0.11569633753970265,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.1553788185119629,
+      "learning_rate": 4.3482261692267186e-07,
+      "loss": 0.05866732001304627,
+      "mean_token_accuracy": 0.9790047742426395,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.23489533165513082,
+      "eval_loss": 0.9410145878791809,
+      "eval_mean_token_accuracy": 0.8524025068726651,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 85.5221,
+      "eval_samples_per_second": 16.078,
+      "eval_steps_per_second": 2.011,
+      "step": 3920
+    },
+    {
+      "entropy": 0.12030278495512903,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.20454250276088715,
+      "learning_rate": 2.797298117403634e-07,
+      "loss": 0.061210107803344724,
+      "mean_token_accuracy": 0.9774218738079071,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.2348609700105911,
+      "eval_loss": 0.9410302639007568,
+      "eval_mean_token_accuracy": 0.8523409498292346,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 85.9105,
+      "eval_samples_per_second": 16.005,
+      "eval_steps_per_second": 2.002,
+      "step": 3940
+    },
+    {
+      "entropy": 0.11162231937050819,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.13079790771007538,
+      "learning_rate": 1.5867404281932237e-07,
+      "loss": 0.05349419116973877,
+      "mean_token_accuracy": 0.9791230395436287,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.23484029950097549,
+      "eval_loss": 0.9411523342132568,
+      "eval_mean_token_accuracy": 0.8523351938225502,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 85.1298,
+      "eval_samples_per_second": 16.152,
+      "eval_steps_per_second": 2.02,
+      "step": 3960
+    },
+    {
+      "entropy": 0.12446248792111873,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705058217048645,
+      "learning_rate": 7.16918189283218e-08,
+      "loss": 0.061916273832321164,
+      "mean_token_accuracy": 0.9767685994505882,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.2347967088395773,
+      "eval_loss": 0.9411665201187134,
+      "eval_mean_token_accuracy": 0.8524297087691551,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.2026,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 3980
+    },
+    {
+      "entropy": 0.13053828622214497,
+      "epoch": 9.951432129514322,
+      "grad_norm": 0.1882583051919937,
+      "learning_rate": 1.8809372719714117e-08,
+      "loss": 0.06384563446044922,
+      "mean_token_accuracy": 0.9753879077732563,
+      "num_tokens": 9341118.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.951432129514322,
+      "eval_entropy": 0.23490050339768098,
+      "eval_loss": 0.9411982297897339,
+      "eval_mean_token_accuracy": 0.8522821709860203,
+      "eval_num_tokens": 9341118.0,
+      "eval_runtime": 86.5264,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 4000
+    },
+    {
+      "entropy": 0.1327908534843188,
+      "epoch": 10.0,
+      "grad_norm": 0.29403096437454224,
+      "learning_rate": 4.2652817989878134e-11,
+      "loss": 0.059251207113265994,
+      "mean_token_accuracy": 0.9757734407217075,
+      "num_tokens": 9386580.0,
+      "step": 4020
+    },
+    {
+      "epoch": 10.0,
+      "eval_entropy": 0.23489686595492584,
+      "eval_loss": 0.9411709904670715,
+      "eval_mean_token_accuracy": 0.8524181870527046,
+      "eval_num_tokens": 9386580.0,
+      "eval_runtime": 86.0874,
+      "eval_samples_per_second": 15.972,
+      "eval_steps_per_second": 1.998,
+      "step": 4020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.964083844973138e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0a3d4e6feec03c88f56611acd0f157a83cecd23
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-420/trainer_state.json
@@ -0,0 +1,475 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0448318804483188,
+  "eval_steps": 20,
+  "global_step": 420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.155416952633139e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a242b67a0860cfea46ff4b66f1b3d40a95a5ec44
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-440/trainer_state.json
@@ -0,0 +1,496 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0946450809464507,
+  "eval_steps": 20,
+  "global_step": 440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.352373181726925e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..eff97ae7ce361c37a9d856c9d0a9610a0689b070
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-460/trainer_state.json
@@ -0,0 +1,517 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1444582814445827,
+  "eval_steps": 20,
+  "global_step": 460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.572855012445594e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6f4baf01ccf7c7ae92c77395ef4ced08db8968a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-480/trainer_state.json
@@ -0,0 +1,538 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1942714819427147,
+  "eval_steps": 20,
+  "global_step": 480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.77945126296064e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2133c2164c1e9050872ce5895ff8f211776ef11f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-500/trainer_state.json
@@ -0,0 +1,559 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.244084682440847,
+  "eval_steps": 20,
+  "global_step": 500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.964191412767334e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..37d8033c2ad3a80d2c22f5bb3f8bb5b452760a76
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-520/trainer_state.json
@@ -0,0 +1,580 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.293897882938979,
+  "eval_steps": 20,
+  "global_step": 520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.160321867214234e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0227ba6fa0308d5ea09a21d1ad046e27fa86b7b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-540/trainer_state.json
@@ -0,0 +1,601 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3437110834371109,
+  "eval_steps": 20,
+  "global_step": 540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.344020822900941e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..250779e39b5be85845e766c9abddb99f0d3563e5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-560/trainer_state.json
@@ -0,0 +1,622 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3935242839352429,
+  "eval_steps": 20,
+  "global_step": 560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.543921118127104e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa06579e5168d43db1d5ea2fd207708bc88c4940
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-580/trainer_state.json
@@ -0,0 +1,643 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4433374844333748,
+  "eval_steps": 20,
+  "global_step": 580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.739225797927117e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b37bba4dec5eda89a0974d894feebb1840d1a341
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-60/trainer_state.json
@@ -0,0 +1,97 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.149439601494396,
+  "eval_steps": 20,
+  "global_step": 60,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5653773829644288.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a1980ad266266897a7926ab36e4ece2f57301c2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-600/trainer_state.json
@@ -0,0 +1,664 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4931506849315068,
+  "eval_steps": 20,
+  "global_step": 600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.93531137331712e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7ad7873298c0936f72d44a8c30749914d3b6c30
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-620/trainer_state.json
@@ -0,0 +1,685 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5429638854296388,
+  "eval_steps": 20,
+  "global_step": 620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.147230279980032e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbb57807b9fb3c69aebdedd79e33236979915ca8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-640/trainer_state.json
@@ -0,0 +1,706 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.592777085927771,
+  "eval_steps": 20,
+  "global_step": 640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.338594578584576e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ae83ae83d0886071cab45a575474461870b6191
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-660/trainer_state.json
@@ -0,0 +1,727 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.6425902864259028,
+  "eval_steps": 20,
+  "global_step": 660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.516450281063424e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..988d360a2d02fb6a3da0a0590161c4c2e57bfef7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-680/trainer_state.json
@@ -0,0 +1,748 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.692403486924035,
+  "eval_steps": 20,
+  "global_step": 680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.728503824897024e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..41394ac4edb51ad7d42d375feef63728b399e67b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-700/trainer_state.json
@@ -0,0 +1,769 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.7422166874221667,
+  "eval_steps": 20,
+  "global_step": 700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.93499236567552e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ee32bba65ed1d5f91fef859c85d69f8a8a47140
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-720/trainer_state.json
@@ -0,0 +1,790 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.792029887920299,
+  "eval_steps": 20,
+  "global_step": 720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.120576241751859e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e707c52c7c8f77754359ded2aed88e806c1782a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-740/trainer_state.json
@@ -0,0 +1,811 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.841843088418431,
+  "eval_steps": 20,
+  "global_step": 740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.32430920843694e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2913162aa55ec3cf060cff741c07c94d7bbae36
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-760/trainer_state.json
@@ -0,0 +1,832 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.891656288916563,
+  "eval_steps": 20,
+  "global_step": 760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.507783768839168e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..020d2663dc76d96cd13dc39a90a5578eff6cf297
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-780/trainer_state.json
@@ -0,0 +1,853 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9414694894146949,
+  "eval_steps": 20,
+  "global_step": 780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.706427450472243e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f69e16ee9b7e91bf931d4d1a703bd2a30318af2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-80/trainer_state.json
@@ -0,0 +1,118 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.199252801992528,
+  "eval_steps": 20,
+  "global_step": 80,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7798902991159296.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4692d06fe7299372f766ab8095d62e7b9f0e06d5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-800/trainer_state.json
@@ -0,0 +1,874 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9912826899128269,
+  "eval_steps": 20,
+  "global_step": 800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.897684039340237e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..975de21d9beed17925bae8047e0658476e033855
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-820/trainer_state.json
@@ -0,0 +1,895 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0398505603985058,
+  "eval_steps": 20,
+  "global_step": 820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.08591129186775e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..30c69ef80eb9fcf466ce3b39d1336bc4e7bd7ced
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-840/trainer_state.json
@@ -0,0 +1,916 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0896637608966375,
+  "eval_steps": 20,
+  "global_step": 840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.296977496449638e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6bedf650e993099d83d679fce5d23b944c11169b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-860/trainer_state.json
@@ -0,0 +1,937 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1394769613947697,
+  "eval_steps": 20,
+  "global_step": 860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.478540209028096e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc2f41cfb549a33778ae4439584a8937d7530bf9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-880/trainer_state.json
@@ -0,0 +1,958 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1892901618929015,
+  "eval_steps": 20,
+  "global_step": 880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.682066732051456e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf4ab10a1faee94af32c354905aed625a93cc42b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-900/trainer_state.json
@@ -0,0 +1,979 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2391033623910337,
+  "eval_steps": 20,
+  "global_step": 900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.875351854291149e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..02c861e2f86988ed4546111ad25433ea8b457c55
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-920/trainer_state.json
@@ -0,0 +1,1000 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2889165628891655,
+  "eval_steps": 20,
+  "global_step": 920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.073672406714573e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6654eefea15fe79488284643fec786dac2bff768
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-940/trainer_state.json
@@ -0,0 +1,1021 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3387297633872977,
+  "eval_steps": 20,
+  "global_step": 940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.248826389968282e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f587407e6cf75c66dc09c00848c0b2bfd3e5d353
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-960/trainer_state.json
@@ -0,0 +1,1042 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3885429638854294,
+  "eval_steps": 20,
+  "global_step": 960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.454911019234714e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e1911440bbdf4e34c1f99d492253e23cdedc8b2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.028265386974777595,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0c4619e989c05ec7b27f97335cf33a4d1ba8645
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-980/trainer_state.json
@@ -0,0 +1,1063 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4383561643835616,
+  "eval_steps": 20,
+  "global_step": 980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.955029806494713,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.020533561706543,
+      "learning_rate": 1.0694800202775147e-05,
+      "loss": 1.7107986450195312,
+      "mean_token_accuracy": 0.6487608112394809,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.3144892034835594,
+      "eval_loss": 1.1198534965515137,
+      "eval_mean_token_accuracy": 0.7460246955932572,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.0565,
+      "eval_samples_per_second": 15.794,
+      "eval_steps_per_second": 1.976,
+      "step": 20
+    },
+    {
+      "entropy": 1.0063214391469955,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.572906494140625,
+      "learning_rate": 2.1952484626748985e-05,
+      "loss": 0.8663722991943359,
+      "mean_token_accuracy": 0.7779282338917255,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7921617945959402,
+      "eval_loss": 0.7062025666236877,
+      "eval_mean_token_accuracy": 0.8100443180910376,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.5189,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 40
+    },
+    {
+      "entropy": 0.7682028576731682,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3003711700439453,
+      "learning_rate": 3.3210169050722824e-05,
+      "loss": 0.673183822631836,
+      "mean_token_accuracy": 0.8182129614055157,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7059133584762729,
+      "eval_loss": 0.6481946706771851,
+      "eval_mean_token_accuracy": 0.8227418761613757,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.5098,
+      "eval_samples_per_second": 15.894,
+      "eval_steps_per_second": 1.988,
+      "step": 60
+    },
+    {
+      "entropy": 0.7029960259795189,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.3664201498031616,
+      "learning_rate": 4.4467853474696664e-05,
+      "loss": 0.6354611873626709,
+      "mean_token_accuracy": 0.8243416830897331,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6867308004997498,
+      "eval_loss": 0.6179807186126709,
+      "eval_mean_token_accuracy": 0.8302594811417335,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.3969,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.991,
+      "step": 80
+    },
+    {
+      "entropy": 0.6764581337571144,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9815880656242371,
+      "learning_rate": 5.57255378986705e-05,
+      "loss": 0.5988658905029297,
+      "mean_token_accuracy": 0.8329168625175953,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6790881479202315,
+      "eval_loss": 0.5998476147651672,
+      "eval_mean_token_accuracy": 0.8318756420251935,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.6653,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 100
+    },
+    {
+      "entropy": 0.6744543805718421,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.932099461555481,
+      "learning_rate": 6.698322232264434e-05,
+      "loss": 0.5991750717163086,
+      "mean_token_accuracy": 0.8304223112761975,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.6813044282932614,
+      "eval_loss": 0.5922021269798279,
+      "eval_mean_token_accuracy": 0.8346439617317777,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.1551,
+      "eval_samples_per_second": 15.96,
+      "eval_steps_per_second": 1.996,
+      "step": 120
+    },
+    {
+      "entropy": 0.6663189359009266,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9528499841690063,
+      "learning_rate": 7.824090674661818e-05,
+      "loss": 0.5891091346740722,
+      "mean_token_accuracy": 0.832152470946312,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6398407208711602,
+      "eval_loss": 0.5859636664390564,
+      "eval_mean_token_accuracy": 0.8372074996316156,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.2706,
+      "eval_samples_per_second": 15.938,
+      "eval_steps_per_second": 1.994,
+      "step": 140
+    },
+    {
+      "entropy": 0.64859763905406,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8468204140663147,
+      "learning_rate": 8.949859117059201e-05,
+      "loss": 0.569426441192627,
+      "mean_token_accuracy": 0.8401990942656994,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6381674285891444,
+      "eval_loss": 0.5744525790214539,
+      "eval_mean_token_accuracy": 0.838626817908398,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.2848,
+      "eval_samples_per_second": 15.936,
+      "eval_steps_per_second": 1.993,
+      "step": 160
+    },
+    {
+      "entropy": 0.6432608783245086,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8765804767608643,
+      "learning_rate": 0.00010075627559456587,
+      "loss": 0.5687318801879883,
+      "mean_token_accuracy": 0.839249350130558,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6047098288355872,
+      "eval_loss": 0.5679298043251038,
+      "eval_mean_token_accuracy": 0.8410577181466791,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.5879,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.986,
+      "step": 180
+    },
+    {
+      "entropy": 0.6440276011824608,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9576020240783691,
+      "learning_rate": 0.00011201396001853971,
+      "loss": 0.5828506469726562,
+      "mean_token_accuracy": 0.837553184479475,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6161119360909906,
+      "eval_loss": 0.5702911615371704,
+      "eval_mean_token_accuracy": 0.8407089398350827,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3341,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 200
+    },
+    {
+      "entropy": 0.6377195850014686,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7212373614311218,
+      "learning_rate": 0.00012327164444251353,
+      "loss": 0.5702451229095459,
+      "mean_token_accuracy": 0.8397969007492065,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6080108886194784,
+      "eval_loss": 0.5633499622344971,
+      "eval_mean_token_accuracy": 0.8396634854549585,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.4945,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.989,
+      "step": 220
+    },
+    {
+      "entropy": 0.6287345830351114,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.848779022693634,
+      "learning_rate": 0.00013452932886648739,
+      "loss": 0.5506546020507812,
+      "mean_token_accuracy": 0.8438881888985634,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6307531505130058,
+      "eval_loss": 0.5573338270187378,
+      "eval_mean_token_accuracy": 0.8431362606758295,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.3535,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.992,
+      "step": 240
+    },
+    {
+      "entropy": 0.6223786748945713,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.7316951751708984,
+      "learning_rate": 0.0001457870132904612,
+      "loss": 0.5495625972747803,
+      "mean_token_accuracy": 0.8440376669168472,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.623454462476941,
+      "eval_loss": 0.5619264245033264,
+      "eval_mean_token_accuracy": 0.8431175777385401,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.2008,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.995,
+      "step": 260
+    },
+    {
+      "entropy": 0.6281675305217505,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7639564871788025,
+      "learning_rate": 0.00015704469771443506,
+      "loss": 0.5604369163513183,
+      "mean_token_accuracy": 0.8401600055396556,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.63416675980701,
+      "eval_loss": 0.5612760782241821,
+      "eval_mean_token_accuracy": 0.842435666294985,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.25,
+      "eval_samples_per_second": 15.942,
+      "eval_steps_per_second": 1.994,
+      "step": 280
+    },
+    {
+      "entropy": 0.6427909277379513,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6475813388824463,
+      "learning_rate": 0.0001683023821384089,
+      "loss": 0.573763370513916,
+      "mean_token_accuracy": 0.8370340794324875,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6231539840268534,
+      "eval_loss": 0.5566866397857666,
+      "eval_mean_token_accuracy": 0.844177934319474,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.4858,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.989,
+      "step": 300
+    },
+    {
+      "entropy": 0.6226776849478484,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.8886699676513672,
+      "learning_rate": 0.00017956006656238274,
+      "loss": 0.558210802078247,
+      "mean_token_accuracy": 0.84083157107234,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6066981683983359,
+      "eval_loss": 0.5585207939147949,
+      "eval_mean_token_accuracy": 0.8423153311014175,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.3463,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 320
+    },
+    {
+      "entropy": 0.6249004438519478,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8791211843490601,
+      "learning_rate": 0.00019081775098635657,
+      "loss": 0.5603597164154053,
+      "mean_token_accuracy": 0.8420463085174561,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6082247584018596,
+      "eval_loss": 0.5616299510002136,
+      "eval_mean_token_accuracy": 0.8431286801432454,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.1253,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.997,
+      "step": 340
+    },
+    {
+      "entropy": 0.6362396612763405,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.8606319427490234,
+      "learning_rate": 0.0002020754354103304,
+      "loss": 0.5735773563385009,
+      "mean_token_accuracy": 0.8371490836143494,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6492362072648004,
+      "eval_loss": 0.5646467804908752,
+      "eval_mean_token_accuracy": 0.8415517574825953,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.3351,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 360
+    },
+    {
+      "entropy": 0.638665035739541,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.7773950099945068,
+      "learning_rate": 0.00021333311983430425,
+      "loss": 0.5820859909057617,
+      "mean_token_accuracy": 0.8372561208903789,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6434498637221581,
+      "eval_loss": 0.5645168423652649,
+      "eval_mean_token_accuracy": 0.8420382481674815,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.1216,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 380
+    },
+    {
+      "entropy": 0.6316851265728474,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 1.6120579242706299,
+      "learning_rate": 0.00022459080425827807,
+      "loss": 0.5637502670288086,
+      "mean_token_accuracy": 0.8386227294802666,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6469012776086497,
+      "eval_loss": 0.5758090615272522,
+      "eval_mean_token_accuracy": 0.8397158470957778,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.6139,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.986,
+      "step": 400
+    },
+    {
+      "entropy": 0.5894816922835815,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 1.1616325378417969,
+      "learning_rate": 0.00022626713048053178,
+      "loss": 0.5316025257110596,
+      "mean_token_accuracy": 0.8466163017810919,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5860798164855602,
+      "eval_loss": 0.5777581930160522,
+      "eval_mean_token_accuracy": 0.8396938103576039,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.1449,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.997,
+      "step": 420
+    },
+    {
+      "entropy": 0.5818420693278312,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.7999453544616699,
+      "learning_rate": 0.00022622107023288778,
+      "loss": 0.5221010208129883,
+      "mean_token_accuracy": 0.8474301159381866,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5783926014636838,
+      "eval_loss": 0.5700300931930542,
+      "eval_mean_token_accuracy": 0.8430753537388735,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.5308,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.988,
+      "step": 440
+    },
+    {
+      "entropy": 0.5612493887543678,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 1.015687346458435,
+      "learning_rate": 0.00022614090619491568,
+      "loss": 0.5084867000579834,
+      "mean_token_accuracy": 0.8495561093091964,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5841563874205877,
+      "eval_loss": 0.5693665742874146,
+      "eval_mean_token_accuracy": 0.8427817298229351,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.5256,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.988,
+      "step": 460
+    },
+    {
+      "entropy": 0.5828216474503278,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 1.9750930070877075,
+      "learning_rate": 0.00022602666254299594,
+      "loss": 0.5180017948150635,
+      "mean_token_accuracy": 0.8515685826539994,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5806607044366903,
+      "eval_loss": 0.5804352760314941,
+      "eval_mean_token_accuracy": 0.8413014668364858,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.1199,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.997,
+      "step": 480
+    },
+    {
+      "entropy": 0.5926914308220148,
+      "epoch": 1.244084682440847,
+      "grad_norm": 0.8917353749275208,
+      "learning_rate": 0.0002258783737314558,
+      "loss": 0.528910779953003,
+      "mean_token_accuracy": 0.8486074328422546,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5593361884009006,
+      "eval_loss": 0.5675153732299805,
+      "eval_mean_token_accuracy": 0.8433507802181466,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7289,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5865630559623242,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7482362985610962,
+      "learning_rate": 0.00022569608448217823,
+      "loss": 0.5250466823577881,
+      "mean_token_accuracy": 0.8477916084229946,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.543057840230853,
+      "eval_loss": 0.5671008229255676,
+      "eval_mean_token_accuracy": 0.8428726016088973,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.3403,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.992,
+      "step": 520
+    },
+    {
+      "entropy": 0.5870206747204065,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9473814964294434,
+      "learning_rate": 0.00022547984977111448,
+      "loss": 0.5252370834350586,
+      "mean_token_accuracy": 0.8468369916081429,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.590982622878496,
+      "eval_loss": 0.5676343441009521,
+      "eval_mean_token_accuracy": 0.8429348746011424,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.5168,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.988,
+      "step": 540
+    },
+    {
+      "entropy": 0.5785854265093804,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.9353351593017578,
+      "learning_rate": 0.0002252297348117042,
+      "loss": 0.5304938316345215,
+      "mean_token_accuracy": 0.8463383808732032,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6099918867612995,
+      "eval_loss": 0.5620437860488892,
+      "eval_mean_token_accuracy": 0.8430728347495545,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.7741,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.982,
+      "step": 560
+    },
+    {
+      "entropy": 0.5768801040947438,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.9198738932609558,
+      "learning_rate": 0.0002249458150352077,
+      "loss": 0.520513391494751,
+      "mean_token_accuracy": 0.8487689301371575,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.6349420670506566,
+      "eval_loss": 0.5645340085029602,
+      "eval_mean_token_accuracy": 0.8447844597489335,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.3257,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.992,
+      "step": 580
+    },
+    {
+      "entropy": 0.5822233572602272,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.832811176776886,
+      "learning_rate": 0.0002246281760679571,
+      "loss": 0.5295282363891601,
+      "mean_token_accuracy": 0.8504064798355102,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5829724387027496,
+      "eval_loss": 0.5612193942070007,
+      "eval_mean_token_accuracy": 0.8449643853791925,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6617,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.571855777129531,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.7665547728538513,
+      "learning_rate": 0.00022427691370553263,
+      "loss": 0.5187931060791016,
+      "mean_token_accuracy": 0.8534420043230057,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5623592240519302,
+      "eval_loss": 0.5575760006904602,
+      "eval_mean_token_accuracy": 0.8468210229346919,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.6324,
+      "eval_samples_per_second": 15.872,
+      "eval_steps_per_second": 1.985,
+      "step": 620
+    },
+    {
+      "entropy": 0.5740394659340382,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6500429511070251,
+      "learning_rate": 0.00022389213388387174,
+      "loss": 0.5283198833465577,
+      "mean_token_accuracy": 0.8502798482775689,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5548852207355721,
+      "eval_loss": 0.5561797022819519,
+      "eval_mean_token_accuracy": 0.8452786498291548,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.5205,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.988,
+      "step": 640
+    },
+    {
+      "entropy": 0.6020145989954472,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.7056867480278015,
+      "learning_rate": 0.00022347395264732053,
+      "loss": 0.5400049209594726,
+      "mean_token_accuracy": 0.8447613954544068,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5618055154417836,
+      "eval_loss": 0.556106686592102,
+      "eval_mean_token_accuracy": 0.8465680112672407,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.2971,
+      "eval_samples_per_second": 15.933,
+      "eval_steps_per_second": 1.993,
+      "step": 660
+    },
+    {
+      "entropy": 0.5665927153080702,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.5987663865089417,
+      "learning_rate": 0.00022302249611363625,
+      "loss": 0.5143643856048584,
+      "mean_token_accuracy": 0.8529589556157589,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.568248552118623,
+      "eval_loss": 0.5476346015930176,
+      "eval_mean_token_accuracy": 0.8476775434128073,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.9583,
+      "eval_samples_per_second": 15.812,
+      "eval_steps_per_second": 1.978,
+      "step": 680
+    },
+    {
+      "entropy": 0.5673687808215618,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.735261857509613,
+      "learning_rate": 0.00022253790043595193,
+      "loss": 0.509885597229004,
+      "mean_token_accuracy": 0.8537046857178211,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5616967284748721,
+      "eval_loss": 0.5439274311065674,
+      "eval_mean_token_accuracy": 0.8488946217437123,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.0604,
+      "eval_samples_per_second": 15.977,
+      "eval_steps_per_second": 1.999,
+      "step": 700
+    },
+    {
+      "entropy": 0.5529541682451964,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7014835476875305,
+      "learning_rate": 0.00022202031176171442,
+      "loss": 0.5078992366790771,
+      "mean_token_accuracy": 0.8525233261287213,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5827173320359962,
+      "eval_loss": 0.5419450402259827,
+      "eval_mean_token_accuracy": 0.8477318609176681,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 85.2984,
+      "eval_samples_per_second": 16.12,
+      "eval_steps_per_second": 2.016,
+      "step": 720
+    },
+    {
+      "entropy": 0.5755720350891351,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.705613911151886,
+      "learning_rate": 0.00022146988618860824,
+      "loss": 0.5181350708007812,
+      "mean_token_accuracy": 0.8467609457671642,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5743971356125765,
+      "eval_loss": 0.5415896773338318,
+      "eval_mean_token_accuracy": 0.847328585940738,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 85.5602,
+      "eval_samples_per_second": 16.071,
+      "eval_steps_per_second": 2.01,
+      "step": 740
+    },
+    {
+      "entropy": 0.561330484598875,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.6722865700721741,
+      "learning_rate": 0.0002208867897174789,
+      "loss": 0.499837589263916,
+      "mean_token_accuracy": 0.8518734864890576,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.5865232653396074,
+      "eval_loss": 0.5437926650047302,
+      "eval_mean_token_accuracy": 0.8450997017843779,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4116,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.547389242425561,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.7935577034950256,
+      "learning_rate": 0.00022027119820226907,
+      "loss": 0.4977591514587402,
+      "mean_token_accuracy": 0.8539491161704064,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.5290903090391048,
+      "eval_loss": 0.5409526824951172,
+      "eval_mean_token_accuracy": 0.8497545698354411,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.7262,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.983,
+      "step": 780
+    },
+    {
+      "entropy": 0.5687909748405218,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6180546283721924,
+      "learning_rate": 0.00021962329729698345,
+      "loss": 0.5109643459320068,
+      "mean_token_accuracy": 0.8521598495543004,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.5503541858390321,
+      "eval_loss": 0.5361555218696594,
+      "eval_mean_token_accuracy": 0.8510884285666221,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.3339,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 800
+    },
+    {
+      "entropy": 0.4739728841261986,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.8058829307556152,
+      "learning_rate": 0.0002189432823996982,
+      "loss": 0.4204097747802734,
+      "mean_token_accuracy": 0.8728981889211215,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.5077334992414297,
+      "eval_loss": 0.5531114339828491,
+      "eval_mean_token_accuracy": 0.8489257208136625,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.4801,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.989,
+      "step": 820
+    },
+    {
+      "entropy": 0.4594309840351343,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.6906896829605103,
+      "learning_rate": 0.0002182313585936314,
+      "loss": 0.4071959495544434,
+      "mean_token_accuracy": 0.8732857562601566,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.49850136994622474,
+      "eval_loss": 0.5486204624176025,
+      "eval_mean_token_accuracy": 0.8507991450470548,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.3364,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.992,
+      "step": 840
+    },
+    {
+      "entropy": 0.4881629109382629,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6343470215797424,
+      "learning_rate": 0.0002174877405852928,
+      "loss": 0.41669540405273436,
+      "mean_token_accuracy": 0.8711295068264008,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.49155513924914734,
+      "eval_loss": 0.555109441280365,
+      "eval_mean_token_accuracy": 0.8496399400539176,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 86.3295,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.992,
+      "step": 860
+    },
+    {
+      "entropy": 0.4648668970912695,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.8014165163040161,
+      "learning_rate": 0.00021671265263973133,
+      "loss": 0.4110250473022461,
+      "mean_token_accuracy": 0.8754166305065155,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.4909258722219356,
+      "eval_loss": 0.5539511442184448,
+      "eval_mean_token_accuracy": 0.8492401502160138,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.3468,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.992,
+      "step": 880
+    },
+    {
+      "entropy": 0.4824485514312983,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6665191054344177,
+      "learning_rate": 0.00021590632851289967,
+      "loss": 0.4181404113769531,
+      "mean_token_accuracy": 0.8726993151009083,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.4986876940657926,
+      "eval_loss": 0.547695517539978,
+      "eval_mean_token_accuracy": 0.8501384708770486,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.3838,
+      "eval_samples_per_second": 15.917,
+      "eval_steps_per_second": 1.991,
+      "step": 900
+    },
+    {
+      "entropy": 0.4751896943897009,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.81158047914505,
+      "learning_rate": 0.00021506901138115678,
+      "loss": 0.40689678192138673,
+      "mean_token_accuracy": 0.8745221219956875,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.507153491121392,
+      "eval_loss": 0.5501641631126404,
+      "eval_mean_token_accuracy": 0.8495670116918032,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.0912,
+      "eval_samples_per_second": 15.971,
+      "eval_steps_per_second": 1.998,
+      "step": 920
+    },
+    {
+      "entropy": 0.4873133715242147,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7218056321144104,
+      "learning_rate": 0.0002142009537679292,
+      "loss": 0.42701358795166017,
+      "mean_token_accuracy": 0.8695114746689796,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5202612736543943,
+      "eval_loss": 0.5491839051246643,
+      "eval_mean_token_accuracy": 0.8494071208460386,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.1142,
+      "eval_samples_per_second": 15.967,
+      "eval_steps_per_second": 1.997,
+      "step": 940
+    },
+    {
+      "entropy": 0.4762951169162989,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 0.7194424867630005,
+      "learning_rate": 0.0002133024174675534,
+      "loss": 0.42299847602844237,
+      "mean_token_accuracy": 0.8709790132939815,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4899340462546016,
+      "eval_loss": 0.5522511601448059,
+      "eval_mean_token_accuracy": 0.8492208258357159,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.463,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.989,
+      "step": 960
+    },
+    {
+      "entropy": 0.49650347977876663,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.8406022787094116,
+      "learning_rate": 0.0002123736734663221,
+      "loss": 0.4275330066680908,
+      "mean_token_accuracy": 0.8670595556497573,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.49691385654515996,
+      "eval_loss": 0.5491269826889038,
+      "eval_mean_token_accuracy": 0.850309816210769,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.17,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.996,
+      "step": 980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.654290717400883e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}